CNN with load-save.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-11T03:35:02.732831Z",
     "start_time": "2020-06-11T03:34:57.271716Z"
    }
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import imageio\n",
    "import matplotlib.pyplot as plt\n",
    "import warnings\n",
    "import pandas as pd\n",
    "import cv2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-05-30T19:52:52.438694Z",
     "start_time": "2020-05-30T19:52:47.362613Z"
    },
    "code_folding": [
     0,
     65
    ]
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\ProgramData\\Anaconda3\\lib\\site-packages\\ipykernel_launcher.py:52: UserWarning: No kernel provided, trying to apply Sobel(3, 3).\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "class ImageProcessing:\n",
    "    def __init__(self):\n",
    "        self.readmode = {1 : \"RGB\", 0 : \"Grayscale\"}\n",
    "    \n",
    "    def read_image(self, location = \"\", mode = 1):\n",
    "        \"\"\"\n",
    "            Uses imageio on back.\n",
    "            * location: Directory of image file.\n",
    "            * mode: Image readmode{1 : RGB, 0 : Grayscale}.\n",
    "        \"\"\"\n",
    "        \n",
    "        img = imageio.imread(location)\n",
    "        if mode == 1:\n",
    "            img = img\n",
    "        elif mode == 0:\n",
    "            img = self.convert_color(img, 0)\n",
    "        elif mode == 2:\n",
    "            pass\n",
    "        else:\n",
    "            raise ValueError(f\"Readmode not understood. Choose from {self.readmode}.\")\n",
    "        return img\n",
    "    \n",
    "    def show(self, image, figsize=(5, 5)):\n",
    "        \"\"\"\n",
    "            Uses Matplotlib.pyplot.\n",
    "            * image: A image to be shown.\n",
    "            * figsize: How big image to show. From plt.figure()\n",
    "            \n",
    "        \"\"\"\n",
    "        fig = plt.figure(figsize=figsize)\n",
    "        im = image\n",
    "        plt.imshow(im, cmap='gray')\n",
    "        plt.show()\n",
    "    \n",
    "    def convert_color(self, img, to=0):\n",
    "        if to==0:\n",
    "            return  0.21 * img[:,:,0] + 0.72 * img[:,:,1] + 0.07 * img[:,:,2]\n",
    "        else:\n",
    "            raise ValueError(\"Color conversion can not understood.\")\n",
    "    \n",
    "    def convolve(self, image, kernel = None, padding = \"zero\", stride=(1, 1), show=False, bias = 0):\n",
    "        \"\"\"\n",
    "            * image: A image to be convolved.\n",
    "            * kernel: A filter/window of odd shape for convolution. Used Sobel(3, 3) default.\n",
    "            * padding: Border operation. Available from zero, same, none. \n",
    "            * stride: How frequently do convolution? \n",
    "        \"\"\"\n",
    "        \n",
    "        if len(image.shape) > 3:\n",
    "            raise ValueError(\"Only 2 and 3 channel image supported.\")\n",
    "        if type(kernel) == type(None):\n",
    "            warnings.warn(\"No kernel provided, trying to apply Sobel(3, 3).\")\n",
    "            kernel = np.array([[-1, 0, 1],\n",
    "                              [-1, 0, 1],\n",
    "                              [-1, 0, 1]])\n",
    "            kernel += kernel.T\n",
    "        kshape = kernel.shape\n",
    "        if kshape[0] % 2 != 1 or kshape[1] % 2 != 1:\n",
    "            raise ValueError(\"Please provide odd length of 2d kernel.\")\n",
    "        \n",
    "        if type(stride) == int:\n",
    "                 stride = (stride, stride)\n",
    "        \n",
    "        shape = image.shape\n",
    "        #print(image.shape)\n",
    "        \"\"\"\n",
    "            zeros_h = np.zeros((int(kshape[0]/2), shape[1])).reshape(-1, shape[1])\n",
    "            zeros_v = np.zeros((int(kshape[1]/2), shape[0]+2*int(kshape[1]/2))).reshape(shape[0]+2*int(kshape[1]/2), -1)\n",
    "\n",
    "            h1 = image[0].reshape(-1, shape[1])\n",
    "            h1 = np.repeat(h1, int(kshape[0]/2)).reshape(-1, shape[1])\n",
    "            h2 = image[-1].reshape(-1, shape[1])\n",
    "            h2 = np.repeat(h2, int(kshape[0]/2)).reshape(-1, shape[1])\n",
    "\n",
    "\n",
    "            #zero padding\n",
    "            padded_img = np.vstack((h1, image, h2)) # add rows\n",
    "\n",
    "            pshape = padded_img.shape\n",
    "            v1 = padded_img[:, 0].reshape(pshape[0], -1)\n",
    "            v1 = np.repeat(v1, int(kshape[1]/2)).reshape(pshape[0], -1)\n",
    "            v2 = padded_img[:, -1].reshape(pshape[0], -1)\n",
    "            v2 = np.repeat(v2, int(kshape[1]/2)).reshape(pshape[0], -1)\n",
    "        \"\"\"\n",
    "        if padding == \"zero\":\n",
    "            zeros_h = np.zeros(shape[1]).reshape(-1, shape[1])\n",
    "            zeros_v = np.zeros(shape[0]+2).reshape(shape[0]+2, -1)\n",
    "\n",
    "            #zero padding\n",
    "            padded_img = np.vstack((zeros_h, image, zeros_h)) # add rows\n",
    "            # print(padded_img)\n",
    "            padded_img = np.hstack((zeros_v, padded_img, zeros_v)) # add cols\n",
    "\n",
    "            image = padded_img\n",
    "            shape = image.shape\n",
    "            \n",
    "        elif padding == \"same\":\n",
    "            h1 = image[0].reshape(-1, shape[1])\n",
    "            h2 = image[-1].reshape(-1, shape[1])\n",
    "\n",
    "\n",
    "            #zero padding\n",
    "            padded_img = np.vstack((h1, image, h2)) # add rows\n",
    "\n",
    "            v1 = padded_img[:, 0].reshape(padded_img.shape[0], -1)\n",
    "            v2 = padded_img[:, -1].reshape(padded_img.shape[0], -1)\n",
    "\n",
    "            padded_img = np.hstack((v1, padded_img, v2)) # add cols\n",
    "\n",
    "            image = padded_img\n",
    "            shape = image.shape\n",
    "        elif padding == None:\n",
    "            pass\n",
    "\n",
    "        rv = 0\n",
    "        cimg = []\n",
    "        for r in range(kshape[0], shape[0]+1, stride[0]):\n",
    "            cv = 0\n",
    "            for c in range(kshape[1], shape[1]+1, stride[1]):\n",
    "                chunk = image[rv:r, cv:c]\n",
    "                soma = (np.multiply(chunk, kernel)+bias).sum()\n",
    "                try:\n",
    "                    chunk = int(soma)\n",
    "                except:\n",
    "                    chunk = int(0)\n",
    "                if chunk < 0:\n",
    "                    chunk = 0\n",
    "                if chunk > 255:\n",
    "                    chunk = 255\n",
    "                cimg.append(chunk)\n",
    "                cv+=stride[1]\n",
    "            rv+=stride[0]\n",
    "        cimg = np.array(cimg, dtype=np.uint8).reshape(int(rv/stride[0]), int(cv/stride[1]))\n",
    "        #print((r, c), (rv, cv))\n",
    "        if show:\n",
    "            print(f\"Image convolved with \\nKernel:{kernel}, \\nPadding: {padding}, \\nStride: {stride}\")\n",
    "        return cimg\n",
    "\n",
    "ip = ImageProcessing()\n",
    "img = ip.read_image(\"cb.jpg\", mode=0)\n",
    "#img = np.array([1, 10, 11, 200, 30, 12, 200, 152, 223, 60, 100, \n",
    "                #190, 11, 20, 10, 102, 207, 102, 223, 50, 18, 109, 117, 200, 30]).reshape(5, 5)\n",
    "cv = ip.convolve(img)\n",
    "\n",
    "\n",
    "ip.show(cv)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "code_folding": [
     1,
     22,
     55
    ]
   },
   "outputs": [],
   "source": [
    "# old FFL\n",
    "class FFL():\n",
    "    \"\"\"\n",
    "        A class to create a single input/hidden feedforward layer of given shape.\n",
    "        \n",
    "        Input:\n",
    "        -------\n",
    "        input_shape: Number of input values to layer. If none, tried to check from preious layer.\n",
    "        neurons: Neurons on that layer.\n",
    "        bias: External bias numpy array. If None, used from np.random.randn(neurons)\n",
    "        weights: External weight numpy array. If None, used from np.random.randn(n_input, neurons)\n",
    "        activation: One of [\"relu\", \"sigmoid\", \"tanh\", \"softmax\"].\n",
    "        is_bias: Do we want bias to be used here? Default True.\n",
    "        \n",
    "        Output:\n",
    "        --------\n",
    "        Object of FFL which can be stacked later to use.\n",
    "        \n",
    "        Example:\n",
    "        FFL(2, 2, activation=\"sigmoid\")\n",
    "    \"\"\"\n",
    "    \n",
    "    def __init__(self, input_shape=None, neurons=1, bias=None, weights=None, activation=None, is_bias = True):\n",
    "        \n",
    "        np.random.seed(100)\n",
    "        self.input_shape = input_shape\n",
    "        self.neurons = neurons\n",
    "        self.isbias = is_bias\n",
    "        self.name = \"\"\n",
    "        self.w = weights\n",
    "        if input_shape != None:\n",
    "            self.output_shape = neurons\n",
    "        \n",
    "        \n",
    "        if self.input_shape != None:\n",
    "            self.weights = weights if weights != None else np.random.randn(self.input_shape, neurons)\n",
    "        if(is_bias):\n",
    "            self.bias = bias if bias != None else np.random.randn(neurons)\n",
    "        else:\n",
    "            self.bias = 0\n",
    "            \n",
    "        self.out = None\n",
    "        self.input = None\n",
    "        self.error = None\n",
    "        self.delta = None\n",
    "        activations = [\"relu\", \"sigmoid\", \"tanh\", \"softmax\"]\n",
    "        self.dw = 0\n",
    "        self.db = 0\n",
    "        if activation not in activations and activation != None:\n",
    "             raise ValueError(f\"Activation function not recognised. Use one of {activations} instead.\")\n",
    "        else:\n",
    "            self.activation = activation\n",
    "        self.parameters = np.multiply.reduce(self.weights.shape) + neurons if is_bias == True else 0\n",
    "            \n",
    "           \n",
    "    def activation_dfn(self, r):\n",
    "        \"\"\"\n",
    "            A method of FFL to find derivative of given activation function.\n",
    "        \"\"\"\n",
    "        \n",
    "        \n",
    "        if self.activation is None:\n",
    "            return np.ones(r.shape)\n",
    "        \n",
    "        if self.activation == \"relu\":\n",
    "            r[r<0] = 0\n",
    "            return r\n",
    "\n",
    "        if self.activation == 'tanh':\n",
    "            return 1 - r ** 2\n",
    "\n",
    "        if self.activation == 'sigmoid':\n",
    "            return r * (1 - r)\n",
    "\n",
    "        if self.activation == \"softmax\":\n",
    "            soft = self.activation_fn(r)\n",
    "                    \n",
    "            #s = soft.reshape(-1, 1)\n",
    "            #dsoft = np.diagflat(s) - np.dot(s, s.T)\n",
    "            #diag_soft = dsoft.diagonal()\n",
    "            # take only the diagonal of dsoft i.e i==j only\n",
    "            \"\"\"\n",
    "                soft = a / np.sum(a)\n",
    "\n",
    "                dsoft = np.diag(soft)\n",
    "                for i in range(len(x)):\n",
    "                    for j in range(len(soft)):\n",
    "                        if i == j:\n",
    "                            d = 1\n",
    "                        else:\n",
    "                            d = 0\n",
    "                        dsoft[i][j] = soft[i] * (d - soft[j])\n",
    "            \"\"\"\n",
    "            diag_soft = soft * (1-soft)\n",
    "            return diag_soft\n",
    "        \n",
    "        return r\n",
    "\n",
    "    def activation_fn(self, r):\n",
    "        \"\"\"\n",
    "        A method of FFL which contains the operation and defination of given activation function.\n",
    "        \"\"\"\n",
    "        \n",
    "        if self.activation == None or self.activation == \"linear\":\n",
    "            return r\n",
    "        if self.activation == \"relu\":\n",
    "            r[r<0] = 0\n",
    "            return r\n",
    "        \n",
    "        # tanh\n",
    "        if self.activation == 'tanh':\n",
    "            return np.tanh(r)\n",
    "\n",
    "        # sigmoid\n",
    "        if self.activation == 'sigmoid':\n",
    "            \n",
    "            return 1 / (1 + np.exp(-r))\n",
    "\n",
    "        if self.activation == \"softmax\":\n",
    "            # stable softmax\n",
    "            r = r - np.max(r)\n",
    "            s = np.exp(r)\n",
    "            return s / np.sum(s)\n",
    "        \n",
    "    def apply_activation(self, x):\n",
    "        soma = np.dot(x, self.weights) + self.bias\n",
    "        self.out = self.activation_fn(soma)\n",
    "        \n",
    "        return self.out\n",
    "    def set_n_input(self):\n",
    "#         print(self.input_shape, self.neurons)\n",
    "        self.weights = self.w if self.w != None else np.random.randn(int(self.input_shape), int(self.neurons))        \n",
    "    def set_output_shape(self):\n",
    "        self.set_n_input()\n",
    "        self.get_parameters()\n",
    "        self.output_shape = self.neurons\n",
    "    def get_parameters(self):\n",
    "        self.parameters = self.input_shape *  self.neurons + self.neurons if self.isbias else 0  \n",
    "        return self.parameters\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-11T03:35:12.137641Z",
     "start_time": "2020-06-11T03:35:11.982733Z"
    },
    "code_folding": [
     0,
     1,
     25,
     30,
     46,
     58,
     70,
     102
    ]
   },
   "outputs": [],
   "source": [
    "class Optimizer:\n",
    "    def __init__(self, layers, name=None, learning_rate = 0.01, mr=0.001):\n",
    "        self.name = name\n",
    "        self.learning_rate = learning_rate\n",
    "        self.mr = mr\n",
    "        keys = [\"sgd\", \"iterative\", \"momentum\", \"rmsprop\", \"adagrad\", \"adam\", \"adamax\", \"adadelta\"]\n",
    "        values = [self.sgd, self.iterative, self.momentum, self.rmsprop, self.adagrad, self.adam, self.adamax, self.adadelta]\n",
    "        self.opt_dict = {keys[i]:values[i] for i in range(len(keys))}\n",
    "        if name != None and name in keys:\n",
    "            self.opt_dict[name](layers=layers, training=False)\n",
    "            #pass\n",
    "    def sgd(self, layers, learning_rate=0.01, beta=0.001, training=True):\n",
    "        learning_rate = self.learning_rate\n",
    "        for l in layers:\n",
    "            if l.parameters !=0:\n",
    "                if training:\n",
    "                    l.weights += l.pdelta_weights*self.mr + l.delta_weights * learning_rate\n",
    "                    l.biases += l.pdelta_biases*self.mr + l.delta_biases * learning_rate\n",
    "                    l.pdelta_weights = l.delta_weights\n",
    "                    l.pdelta_biases = l.delta_biases\n",
    "                else:\n",
    "                    l.pdelta_weights = 0\n",
    "                    l.pdelta_biases = 0\n",
    "                    #l.delta_weights = 0\n",
    "                    #l.delta_biases = 0\n",
    "    def iterative(self, layers, learning_rate=0.01, beta=0, training=True):\n",
    "        for l in layers:\n",
    "            if l.parameters !=0:\n",
    "                l.weights -= learning_rate * l.delta_weights\n",
    "                l.biases -= learning_rate * l.delta_biases\n",
    "    def momentum(self, layers, learning_rate=0.1, beta1=0.9, weight_decay=0.0005, nesterov=True, training=True):\n",
    "        learning_rate = self.learning_rate\n",
    "        #beta1 = 1 - self.learning_rate\n",
    "        for l in layers:\n",
    "            if l.parameters !=0:\n",
    "                if training:\n",
    "                    l.weights_momentum = beta1 * l.weights_momentum + learning_rate * l.delta_weights-weight_decay *learning_rate*l.weights\n",
    "                    l.weights+=l.weights_momentum\n",
    "                    #\n",
    "                    l.biases_momentum = beta1 * l.biases_momentum + learning_rate * l.delta_biases-weight_decay *learning_rate*l.biases\n",
    "                    l.biases+=l.biases_momentum\n",
    "                else:\n",
    "                    l.weights_momentum = 0\n",
    "                    l.biases_momentum = 0\n",
    "\n",
    "            \n",
    "    def rmsprop(self, layers, learning_rate=0.001, beta1=0.9, epsilon=1e-8, training=True):\n",
    "        learning_rate=self.learning_rate\n",
    "        for l in layers:\n",
    "            if l.parameters !=0:\n",
    "                if training:\n",
    "                    l.weights_rms = beta1*l.weights_rms + (1-beta1)*(l.delta_weights ** 2)\n",
    "                    l.weights += learning_rate * (l.delta_weights/np.sqrt(l.weights_rms + epsilon))\n",
    "                    l.biases_rms = beta1*l.biases_rms + (1-beta1)*(l.delta_biases ** 2)\n",
    "                    l.biases += learning_rate * (l.delta_biases/np.sqrt(l.biases_rms + epsilon))\n",
    "                else:\n",
    "                    l.weights_rms = 0\n",
    "                    l.biases_rms = 0\n",
    "    def adagrad(self, layers, learning_rate=0.01, beta1=0.9, epsilon=1e-8, training=True):\n",
    "        learning_rate=self.learning_rate\n",
    "        for l in layers:\n",
    "            if l.parameters != 0:\n",
    "                if training:\n",
    "                    l.weights_adagrad += l.delta_weights ** 2\n",
    "                    l.weights += learning_rate * (l.delta_weights/np.sqrt(l.weights_adagrad+epsilon))\n",
    "                    l.biases_adagrad += l.delta_biases ** 2\n",
    "                    l.biases += learning_rate * (l.delta_biases/np.sqrt(l.biases_adagrad+epsilon))\n",
    "                else:\n",
    "                    l.weights_adagrad = 0\n",
    "                    l.biases_adagrad = 0\n",
    "    def adam(self, layers, learning_rate=0.001, beta1=0.9, beta2=0.999, epsilon=1e-8, decay=0, training=True):\n",
    "        #print(training)\n",
    "        for l in layers:\n",
    "            if l.parameters != 0:\n",
    "                if training:\n",
    "                    l.t += 1\n",
    "                    if l.t == 1:\n",
    "                        l.pdelta_biases = 0\n",
    "                        l.pdelta_weights = 0\n",
    "                    l.weights_adam1 = beta1 * l.weights_adam1 + (1-beta1)*l.delta_weights\n",
    "                    l.weights_adam2 = beta2 * l.weights_adam2 + (1-beta2)*(l.delta_weights**2)\n",
    "                    mcap = l.weights_adam1/(1-beta1**l.t)\n",
    "                    vcap = l.weights_adam2/(1-beta2**l.t)\n",
    "                    l.delta_weights = mcap/(np.sqrt(vcap) + epsilon)\n",
    "                    l.weights += l.pdelta_weights * self.mr + learning_rate * l.delta_weights\n",
    "                    l.pdelta_weights = l.delta_weights * 0\n",
    "\n",
    "                    l.biases_adam1 = beta1 * l.biases_adam1 + (1-beta1)*l.delta_biases\n",
    "                    l.biases_adam2 = beta2 * l.biases_adam2 + (1-beta2)*(l.delta_biases**2)\n",
    "                    mcap = l.biases_adam1/(1-beta1**l.t)\n",
    "                    vcap = l.biases_adam2/(1-beta2**l.t)\n",
    "                    l.delta_biases = mcap/(np.sqrt(vcap) +epsilon)\n",
    "                    l.biases += l.pdelta_biases * self.mr + learning_rate * l.delta_biases\n",
    "                    l.pdelta_biases = l.delta_biases * 0\n",
    "                    \n",
    "                else:\n",
    "                    l.t = 0\n",
    "                    l.weights_adam1 = 0\n",
    "                    l.weights_adam2 = 0\n",
    "                    l.biases_adam1 = 0\n",
    "                    l.biases_adam2 = 0\n",
    "                    \n",
    "    def adamax(self, layers, learning_rate=0.002, beta1=0.9, beta2=0.999, epsilon=1e-8, training=True):\n",
    "        for l in layers:\n",
    "            if l.parameters != 0:\n",
    "                if training:\n",
    "                    l.weights_m = beta1*l.weights_m + (1-beta1)*l.delta_weights\n",
    "                    l.weights_v = np.maximum(beta2*l.weights_v, abs(l.delta_weights))\n",
    "                    l.weights += (self.learning_rate/(1-beta1))*(l.weights_m/(l.weights_v+epsilon))\n",
    "                    \n",
    "                    l.biases_m = beta1*l.biases_m + (1-beta1)*l.delta_biases\n",
    "                    l.biases_v = np.maximum(beta2*l.biases_v, abs(l.delta_biases))\n",
    "                    l.biases += (self.learning_rate/(1-beta1))*(l.biases_m/(l.biases_v+epsilon))\n",
    "                    \n",
    "                else:\n",
    "                    l.weights_m = 0\n",
    "                    l.biases_m = 0\n",
    "                    l.weights_v = 0\n",
    "                    l.biases_v = 0\n",
    "                    \n",
    "    def adadelta(self, layers, learning_rate=0.01, beta1=0.9, epsilon=1e-8, training=True):\n",
    "        for l in layers:\n",
    "            if l.parameters != 0:\n",
    "                if training:\n",
    "                    l.weights_v = beta1*l.weights_v + (1-beta1)*(l.delta_weights ** 2)\n",
    "                    l.delta_weights = np.sqrt((l.weights_m+epsilon)/(l.weights_v+epsilon))*l.delta_weights\n",
    "                    l.weights_m = beta1*l.weights_m + (1-beta1)*(l.delta_weights)\n",
    "                    l.weights += l.delta_weights\n",
    "                    \n",
    "                    l.biases_v = beta1*l.biases_v + (1-beta1)*(l.delta_biases ** 2)\n",
    "                    l.delta_biases = np.sqrt((l.biases_m+epsilon)/(l.biases_v+epsilon))*l.delta_biases\n",
    "                    l.biases_m = beta1*l.biases_m+ (1-beta1)*(l.delta_biases)\n",
    "                    l.biases += l.delta_biases\n",
    "                    \n",
    "                else:\n",
    "                    l.weights_m = 0\n",
    "                    l.biases_m = 0\n",
    "                    l.weights_v = 0\n",
    "                    l.biases_v = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-11T03:35:43.170482Z",
     "start_time": "2020-06-11T03:35:21.639061Z"
    },
    "code_folding": [
     5,
     65,
     107,
     129,
     133,
     140,
     155,
     198,
     209,
     226,
     288,
     289,
     317,
     331,
     371,
     419,
     434,
     443,
     445,
     451,
     488,
     491,
     508,
     577,
     618,
     644,
     653,
     665,
     699,
     723,
     761,
     781,
     817,
     836,
     862,
     888,
     975,
     1026,
     1077,
     1128,
     1139,
     1152
    ],
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                      Input Output Shape Activation   Bias  Parameters\n",
      "Layer Name                                                            \n",
      "Input Layer     (28, 28, 1)  (26, 26, 4)       relu   True          40\n",
      "Conv2d1         (26, 26, 4)  (24, 24, 8)       relu   True         296\n",
      "Dropout2        (24, 24, 8)  (24, 24, 8)       None  False           0\n",
      "Upsample3       (24, 24, 8)  (48, 48, 8)       None  False           0\n",
      "Flatten4        (48, 48, 8)        18432       None  False           0\n",
      "Out Layer(FFL)        18432           10    softmax   True      184330\n",
      "Total Parameters: 184666\n",
      "\n",
      "Validation data found.\n",
      "\n",
      "\n",
      "Total 100 samples.\n",
      "Training samples: 100 Validation samples: 5.\n",
      "Total 4 batches, most batch has 32 samples.\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\ProgramData\\Anaconda3\\lib\\site-packages\\ipykernel_launcher.py:815: RuntimeWarning: divide by zero encountered in log\n",
      "ERROR:root:Internal Python error in the inspect module.\n",
      "Below is the traceback from this internal error.\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Traceback (most recent call last):\n",
      "  File \"C:\\ProgramData\\Anaconda3\\lib\\site-packages\\IPython\\core\\interactiveshell.py\", line 3326, in run_code\n",
      "    exec(code_obj, self.user_global_ns, self.user_ns)\n",
      "  File \"<ipython-input-4-5ebc2bf16e67>\", line 1272, in <module>\n",
      "    m.train(x[:100], y[:100], epochs=100, batch_size=32, val_x=xt[:5], val_y=yt[:5])\n",
      "  File \"<ipython-input-4-5ebc2bf16e67>\", line 1202, in train\n",
      "    self.backpropagate(loss, update)\n",
      "  File \"<ipython-input-4-5ebc2bf16e67>\", line 1113, in backpropagate\n",
      "    layer.backpropagate(nx_layer)\n",
      "  File \"<ipython-input-4-5ebc2bf16e67>\", line 281, in backpropagate\n",
      "    layer.delta[i, j, f] = dout\n",
      "ValueError: setting an array element with a sequence.\n",
      "\n",
      "During handling of the above exception, another exception occurred:\n",
      "\n",
      "Traceback (most recent call last):\n",
      "  File \"C:\\ProgramData\\Anaconda3\\lib\\site-packages\\IPython\\core\\interactiveshell.py\", line 2040, in showtraceback\n",
      "    stb = value._render_traceback_()\n",
      "AttributeError: 'ValueError' object has no attribute '_render_traceback_'\n",
      "\n",
      "During handling of the above exception, another exception occurred:\n",
      "\n",
      "Traceback (most recent call last):\n",
      "  File \"C:\\ProgramData\\Anaconda3\\lib\\site-packages\\IPython\\core\\ultratb.py\", line 1101, in get_records\n",
      "    return _fixed_getinnerframes(etb, number_of_lines_of_context, tb_offset)\n",
      "  File \"C:\\ProgramData\\Anaconda3\\lib\\site-packages\\IPython\\core\\ultratb.py\", line 319, in wrapped\n",
      "    return f(*args, **kwargs)\n",
      "  File \"C:\\ProgramData\\Anaconda3\\lib\\site-packages\\IPython\\core\\ultratb.py\", line 353, in _fixed_getinnerframes\n",
      "    records = fix_frame_records_filenames(inspect.getinnerframes(etb, context))\n",
      "  File \"C:\\ProgramData\\Anaconda3\\lib\\inspect.py\", line 1502, in getinnerframes\n",
      "    frameinfo = (tb.tb_frame,) + getframeinfo(tb, context)\n",
      "  File \"C:\\ProgramData\\Anaconda3\\lib\\inspect.py\", line 1460, in getframeinfo\n",
      "    filename = getsourcefile(frame) or getfile(frame)\n",
      "  File \"C:\\ProgramData\\Anaconda3\\lib\\inspect.py\", line 696, in getsourcefile\n",
      "    if getattr(getmodule(object, filename), '__loader__', None) is not None:\n",
      "  File \"C:\\ProgramData\\Anaconda3\\lib\\inspect.py\", line 733, in getmodule\n",
      "    if ismodule(module) and hasattr(module, '__file__'):\n",
      "  File \"C:\\Users\\Quassarian\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow\\__init__.py\", line 50, in __getattr__\n",
      "    module = self._load()\n",
      "  File \"C:\\Users\\Quassarian\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow\\__init__.py\", line 44, in _load\n",
      "    module = _importlib.import_module(self.__name__)\n",
      "  File \"C:\\ProgramData\\Anaconda3\\lib\\importlib\\__init__.py\", line 127, in import_module\n",
      "    return _bootstrap._gcd_import(name[level:], package, level)\n",
      "  File \"<frozen importlib._bootstrap>\", line 1006, in _gcd_import\n",
      "  File \"<frozen importlib._bootstrap>\", line 983, in _find_and_load\n",
      "  File \"<frozen importlib._bootstrap>\", line 967, in _find_and_load_unlocked\n",
      "  File \"<frozen importlib._bootstrap>\", line 677, in _load_unlocked\n",
      "  File \"<frozen importlib._bootstrap_external>\", line 728, in exec_module\n",
      "  File \"<frozen importlib._bootstrap>\", line 219, in _call_with_frames_removed\n",
      "  File \"C:\\Users\\Quassarian\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\contrib\\__init__.py\", line 90, in <module>\n",
      "    from tensorflow.contrib import tensor_forest\n",
      "  File \"C:\\Users\\Quassarian\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\contrib\\tensor_forest\\__init__.py\", line 21, in <module>\n",
      "    from tensorflow.contrib.tensor_forest.client import *\n",
      "  File \"C:\\Users\\Quassarian\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\contrib\\tensor_forest\\__init__.py\", line 21, in <module>\n",
      "    from tensorflow.contrib.tensor_forest.client import *\n",
      "  File \"C:\\Users\\Quassarian\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\contrib\\tensor_forest\\client\\__init__.py\", line 22, in <module>\n",
      "    from tensorflow.contrib.tensor_forest.client import random_forest\n",
      "  File \"C:\\Users\\Quassarian\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\contrib\\tensor_forest\\client\\random_forest.py\", line 27, in <module>\n",
      "    from tensorflow.contrib.tensor_forest.python import tensor_forest\n",
      "  File \"C:\\Users\\Quassarian\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\contrib\\tensor_forest\\python\\__init__.py\", line 21, in <module>\n",
      "    from tensorflow.contrib.tensor_forest.python import tensor_forest\n",
      "  File \"C:\\Users\\Quassarian\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\contrib\\tensor_forest\\python\\tensor_forest.py\", line 31, in <module>\n",
      "    from tensorflow.contrib.tensor_forest.python.ops import stats_ops\n",
      "  File \"C:\\Users\\Quassarian\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\contrib\\tensor_forest\\python\\ops\\stats_ops.py\", line 39, in <module>\n",
      "    resource_loader.get_path_to_datafile(\"_stats_ops.so\"))\n",
      "  File \"C:\\Users\\Quassarian\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\contrib\\util\\loader.py\", line 56, in load_op_library\n",
      "    ret = load_library.load_op_library(path)\n",
      "  File \"C:\\Users\\Quassarian\\AppData\\Roaming\\Python\\Python37\\site-packages\\tensorflow_core\\python\\framework\\load_library.py\", line 61, in load_op_library\n",
      "    lib_handle = py_tf.TF_LoadLibrary(library_filename)\n",
      "KeyboardInterrupt\n"
     ]
    },
    {
     "ename": "ValueError",
     "evalue": "setting an array element with a sequence.",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m"
     ]
    }
   ],
   "source": [
    "# used on blogs\n",
    "#from nn import NN, FFL\n",
    "import time\n",
    "\n",
    "class FFL():\n",
    "    \"\"\"\n",
    "        A class to create a single input/hidden feedforward layer of given shape.\n",
    "        \n",
    "        Input:\n",
    "        -------\n",
    "        n_input: Number of input values to layer. If none, tried to check from preious layer.\n",
    "        neurons: Neurons on that layer.\n",
    "        bias: External bias numpy array. If None, used from np.random.randn(neurons)\n",
    "        weights: External weight numpy array. If None, used from np.random.randn(n_input, neurons)\n",
    "        activation: One of [\"relu\", \"sigmoid\", \"tanh\", \"softmax\"].\n",
    "        is_bias: Do we want bias to be used here? Default True.\n",
    "        \n",
    "        Output:\n",
    "        --------\n",
    "        Object of FFL which can be stacked later to use.\n",
    "        \n",
    "        Example:\n",
    "        FFL(2, 2, activation=\"sigmoid\")\n",
    "    \"\"\"\n",
    "    \n",
    "    def __init__(self, input_shape=None, neurons=1, bias=None, weights=None, activation=None, is_bias = True):\n",
    "        \n",
    "        np.random.seed(100)\n",
    "        self.input_shape = input_shape\n",
    "        self.neurons = neurons\n",
    "        self.isbias = is_bias\n",
    "        self.name = \"\"\n",
    "        \n",
    "        \n",
    "        self.w = weights\n",
    "        self.b = bias\n",
    "        if input_shape != None:\n",
    "            self.output_shape = neurons\n",
    "        \n",
    "        \n",
    "        if self.input_shape != None:\n",
    "            self.weights = weights if weights != None else np.random.randn(self.input_shape, neurons)\n",
    "            self.parameters = self.input_shape *  self.neurons + self.neurons if self.isbias else 0  \n",
    "        if(is_bias):\n",
    "            self.biases = bias if bias != None else np.random.randn(neurons)\n",
    "        else:\n",
    "            self.biases = 0\n",
    "            \n",
    "        self.out = None\n",
    "        self.input = None\n",
    "        self.error = None\n",
    "        self.delta = None\n",
    "        activations = [\"relu\", \"sigmoid\", \"tanh\", \"softmax\"]\n",
    "        self.delta_weights = 0\n",
    "        self.delta_biases = 0\n",
    "        self.pdelta_weights = 0\n",
    "        self.pdelta_biases = 0\n",
    "        \n",
    "        #self.bias = np.ones(self.bias.shape)\n",
    "        \n",
    "        if activation not in activations and activation != None:\n",
    "             raise ValueError(f\"Activation function not recognised. Use one of {activations} instead.\")\n",
    "        else:\n",
    "            self.activation = activation\n",
    "          \n",
    "    def activation_dfn(self, r):\n",
    "        \"\"\"\n",
    "            A method of FFL to find derivative of given activation function.\n",
    "        \"\"\"\n",
    "        if self.activation == \"relu\":\n",
    "            r[r<0] = 0\n",
    "            return r\n",
    "        \n",
    "        if self.activation is None:\n",
    "            return np.ones(r.shape)\n",
    "        \n",
    "        if self.activation == 'tanh':\n",
    "            return 1 - r ** 2\n",
    "\n",
    "        if self.activation == 'sigmoid':\n",
    "            return r * (1 - r)\n",
    "\n",
    "        if self.activation == \"softmax\":\n",
    "            soft = self.activation_fn(r)\n",
    "                    \n",
    "            #s = soft.reshape(-1, 1)\n",
    "            #dsoft = np.diagflat(s) - np.dot(s, s.T)\n",
    "            #diag_soft = dsoft.diagonal()\n",
    "            \n",
    "            # take only the diagonal of dsoft i.e i==j only\n",
    "            \"\"\"\n",
    "                soft = a / np.sum(a)\n",
    "\n",
    "                dsoft = np.diag(soft)\n",
    "                for i in range(len(x)):\n",
    "                    for j in range(len(soft)):\n",
    "                        if i == j:\n",
    "                            d = 1\n",
    "                        else:\n",
    "                            d = 0\n",
    "                        dsoft[i][j] = soft[i] * (d - soft[j])\n",
    "            \"\"\"\n",
    "            diag_soft = soft*(1- soft)\n",
    "            return diag_soft\n",
    "        \n",
    "        return r\n",
    "\n",
    "    def activation_fn(self, r):\n",
    "        \"\"\"\n",
    "        A method of FFL which contains the operation and defination of given activation function.\n",
    "        \"\"\"\n",
    "        \n",
    "        if self.activation == None or self.activation == \"linear\":\n",
    "            return r\n",
    "\n",
    "        # tanh\n",
    "        if self.activation == 'tanh':\n",
    "            return np.tanh(r)\n",
    "\n",
    "        # sigmoid\n",
    "        if self.activation == 'sigmoid':\n",
    "            \n",
    "            return 1 / (1 + np.exp(-r))\n",
    "\n",
    "        if self.activation == \"softmax\":\n",
    "            # stable softmax\n",
    "            r = r - np.max(r)\n",
    "            s = np.exp(r)\n",
    "            return s / np.sum(s)\n",
    "        if self.activation == \"relu\":\n",
    "            r[r<0] = 0\n",
    "            return r\n",
    "        \n",
    "    def apply_activation(self, x):\n",
    "        soma = np.dot(x, self.weights) + self.biases\n",
    "        self.out = self.activation_fn(soma)\n",
    "        \n",
    "        return self.out\n",
    "    def set_n_input(self):\n",
    "        self.weights = self.w if self.w != None else np.random.normal(size=(self.input_shape, self.neurons))\n",
    "    def backpropagate(self, nx_layer):\n",
    "        self.error = np.dot(nx_layer.weights, nx_layer.delta)\n",
    "        self.delta = self.error * self.activation_dfn(self.out)\n",
    "        self.delta_weights += self.delta * np.atleast_2d(self.input).T\n",
    "        self.delta_biases += self.delta\n",
    "    \n",
    "    # added below methods from cnn\n",
    "    def set_output_shape(self):\n",
    "        self.set_n_input()\n",
    "        self.output_shape = self.neurons\n",
    "        self.get_parameters()\n",
    "    def get_parameters(self):\n",
    "        self.parameters = self.input_shape *  self.neurons + self.neurons if self.isbias else 0  \n",
    "        return self.parameters\n",
    "\n",
    "class Dropout:\n",
    "    def __init__(self, prob = 0.5):\n",
    "        self.input_shape=None\n",
    "        self.output_shape = None\n",
    "        self.input_data= None\n",
    "        self.output = None\n",
    "        self.isbias = False\n",
    "        self.activation = None\n",
    "        self.parameters = 0\n",
    "        self.delta = 0\n",
    "        self.weights = 0\n",
    "        self.bias = 0\n",
    "        self.prob = prob\n",
    "        self.delta_weights = 0\n",
    "        self.delta_biases = 0\n",
    "        \n",
    "        \n",
    "    def set_output_shape(self):\n",
    "        #print(self.input_shape)\n",
    "        self.output_shape = self.input_shape\n",
    "        self.weights = 0\n",
    "    def apply_activation(self, x, train=True):\n",
    "        if train:\n",
    "            self.input_data = x\n",
    "            #print(x.sum(axis=2))\n",
    "            flat = np.array(self.input_data).flatten()\n",
    "            random_indices = np.random.randint(0, len(flat), int(self.prob * len(flat)))\n",
    "            flat[random_indices] = 0\n",
    "            self.output = flat.reshape(x.shape)\n",
    "            return self.output\n",
    "        else:\n",
    "            self.input_data = x\n",
    "            self.output = x / self.prob\n",
    "            return self.output\n",
    "    def activation_dfn(self, x):\n",
    "        return x\n",
    "    def backpropagate(self, nx_layer):\n",
    "        if type(nx_layer).__name__ != \"Conv2d\":\n",
    "            self.error = np.dot(nx_layer.weights, nx_layer.delta)\n",
    "            self.delta = self.error * self.activation_dfn(self.out)\n",
    "        else:\n",
    "            self.delta = nx_layer.delta\n",
    "        self.delta[self.output == 0] = 0\n",
    "def one_hot_encoding(lbl, classes):\n",
    "    encoded = np.zeros((len(lbl), classes))\n",
    "    c = list(set(lbl))\n",
    "    if len(c) != classes:\n",
    "        raise ValueError(\"Number of classes is not equal to unique labels.\")\n",
    "    for i in range(len(yy)):\n",
    "        for j in range(len(c)):\n",
    "            if c[j] == lbl[i]:\n",
    "                encoded[i, j] = 1\n",
    "    return encoded\n",
    "class Upsample:\n",
    "    def __init__(self, kernel_size = (2, 2)):\n",
    "        self.input_shape=None\n",
    "        self.output_shape = None\n",
    "        self.input_data= None\n",
    "        self.output = None\n",
    "        self.isbias = False\n",
    "        self.activation = None\n",
    "        self.parameters = 0\n",
    "        self.delta = 0\n",
    "        self.weights = 0\n",
    "        self.bias = 0\n",
    "        self.delta_weights = 0\n",
    "        self.delta_biases = 0        \n",
    "        self.kernel_size = kernel_size\n",
    "        self.stride = self.kernel_size\n",
    "        \n",
    "    \n",
    "    def set_output_shape(self):\n",
    "        shape = self.input_shape\n",
    "        self.output_shape = (shape[0] * self.kernel_size[0], shape[1] * self.kernel_size[1], shape[2])\n",
    "        self.weights = 0\n",
    "    def apply_activation(self, image):\n",
    "        stride = self.stride\n",
    "        kshape = self.kernel_size\n",
    "        self.input_shape = image.shape\n",
    "        self.set_output_shape()\n",
    "        \n",
    "        rstep = stride[0]\n",
    "        cstep = stride[1]\n",
    "        self.out = np.zeros(self.output_shape)\n",
    "        shape = self.output_shape\n",
    "        \n",
    "        \n",
    "        for nc in range(shape[2]):\n",
    "            cimg = []\n",
    "            rv = 0\n",
    "            i = 0\n",
    "            for r in range(kshape[0], shape[0]+1, rstep):\n",
    "                cv = 0\n",
    "                j = 0\n",
    "                for c in range(kshape[1], shape[1]+1, cstep):\n",
    "                    self.out[rv:r, cv:c] = image[i, j]\n",
    "                    j+=1\n",
    "                    cv+=cstep\n",
    "                rv+=rstep\n",
    "                i+=1\n",
    "                \n",
    "        return self.out\n",
    "    def backpropagate(self, nx_layer):\n",
    "        \"\"\"\n",
    "            Gradients are passed through index of largest value .\n",
    "        \"\"\"\n",
    "        layer = self\n",
    "        stride = layer.stride\n",
    "        kshape = layer.kernel_size\n",
    "        image = layer.input\n",
    "        shape = image.shape\n",
    "        layer.delta = np.zeros(shape)\n",
    "        \n",
    "        cimg = []\n",
    "        rstep = stride[0]\n",
    "        cstep = stride[1]\n",
    "        shape = nx_layer.delta.shape\n",
    "        for f in range(shape[2]):\n",
    "            i = 0\n",
    "            rv = 0\n",
    "            for r in range(kshape[0], shape[0]+1, rstep):\n",
    "                cv = 0\n",
    "                j = 0\n",
    "                for c in range(kshape[1], shape[1]+1, cstep):\n",
    "                    dout = nx_layer.delta[rv:r, cv:c, f]\n",
    "                    layer.delta[i, j, f] = dout\n",
    "\n",
    "\n",
    "                    j+=1\n",
    "                    cv+=cstep\n",
    "                rv+=rstep\n",
    "                i+=1\n",
    "            \n",
    "class Pool2d:\n",
    "    def __init__(self, kernel_size = (2, 2), stride=None, kind=\"max\", padding=None):\n",
    "        self.input_shape=None\n",
    "        self.output_shape = None\n",
    "        self.input_data= None\n",
    "        self.output = None\n",
    "        self.isbias = False\n",
    "        self.activation = None\n",
    "        self.parameters = 0\n",
    "        self.delta = 0\n",
    "        self.weights = 0\n",
    "        self.bias = 0\n",
    "        self.delta_weights = 0\n",
    "        self.delta_biases = 0\n",
    "        self.padding = padding\n",
    "        self.p = 1 if padding != None else 0\n",
    "        self.kernel_size = kernel_size\n",
    "        if type(stride) == int:\n",
    "                 stride = (stride, stride)\n",
    "\n",
    "        self.stride = stride\n",
    "        if self.stride == None:\n",
    "            self.stride = self.kernel_size\n",
    "        self.pools = ['max', \"average\", 'min']\n",
    "        if kind not in self.pools:\n",
    "            raise ValueError(\"Pool kind not understoood.\")\n",
    "            \n",
    "        self.kind = kind\n",
    "    \n",
    "    def set_output_shape(self):\n",
    "        shape = self.input_shape\n",
    "        self.output_shape = (int((self.input_shape[0] - self.kernel_size[0] + 2 * self.p) / self.stride[0] + 1), \n",
    "                                int((self.input_shape[1] - self.kernel_size[1] + 2 * self.p) / self.stride[1] + 1), self.input_shape[2])\n",
    "        \n",
    "#         stride = self.stride\n",
    "#         kshape = self.kernel_size\n",
    "#         rstep = stride[0]+kshape[0]-1\n",
    "#         cstep = stride[1]+kshape[1]-1\n",
    "#         rv = rstep * len(range(kshape[0], shape[0]+1, rstep))\n",
    "#         cv = cstep * len(range(kshape[1], shape[1]+1, cstep))\n",
    "#         self.output_shape = (int(rv/rstep), int(cv/cstep), shape[2])\n",
    "#         print(self.output_shape)\n",
    "        self.weights = 0\n",
    "    def apply_activation(self, image):\n",
    "        stride = self.stride\n",
    "        kshape = self.kernel_size\n",
    "        shape = image.shape\n",
    "        self.input_shape = shape\n",
    "        self.set_output_shape()\n",
    "        #print(self.output_shape, shape)\n",
    "        \n",
    "        \n",
    "        rstep = stride[0]+kshape[0]-1\n",
    "        cstep = stride[1]+kshape[1]-1\n",
    "        self.out = np.zeros((self.output_shape))\n",
    "        for nc in range(shape[2]):\n",
    "            cimg = []\n",
    "            rv = 0\n",
    "            for r in range(kshape[0], shape[0]+1, stride[0]):\n",
    "                cv = 0\n",
    "                for c in range(kshape[1], shape[1]+1, stride[1]):\n",
    "                    chunk = image[rv:r, cv:c, nc]\n",
    "                    if len(chunk) > 0:\n",
    "                        \n",
    "                        if self.kind == \"max\":\n",
    "                            #print(chunk)\n",
    "                            chunk = np.max(chunk)\n",
    "                        if self.kind == \"min\":\n",
    "                            chunk = np.min(chunk)\n",
    "                        if self.kind == \"average\":\n",
    "                            chunk = np.mean(chunk)\n",
    "\n",
    "                        cimg.append(chunk)\n",
    "                    else:\n",
    "                        cv-=cstep\n",
    "                        #rv-=rstep\n",
    "                    cv+=stride[1]\n",
    "                rv+=stride[0]\n",
    "            cimg = np.array(cimg).reshape(int(rv/stride[0]), int(cv/stride[1]))\n",
    "            #print(cimg.shape)\n",
    "            self.out[:,:,nc] = cimg\n",
    "        return self.out\n",
    "    def backpropagate(self, nx_layer):\n",
    "        \"\"\"\n",
    "            Gradients are passed through index of largest value .\n",
    "        \"\"\"\n",
    "        layer = self\n",
    "        stride = layer.stride\n",
    "        kshape = layer.kernel_size\n",
    "        image = layer.input\n",
    "        shape = image.shape\n",
    "        layer.delta = np.zeros(shape)\n",
    "        #self.input_shape = shape\n",
    "        #self.set_output_shape()\n",
    "        #print(self.output_shape)\n",
    "        \n",
    "        cimg = []\n",
    "        rstep = stride[0]\n",
    "        cstep = stride[1]\n",
    "        \n",
    "        for f in range(shape[2]):\n",
    "            i = 0\n",
    "            rv = 0\n",
    "            for r in range(kshape[0], shape[0]+1, rstep):\n",
    "                cv = 0\n",
    "                j = 0\n",
    "                for c in range(kshape[1], shape[1]+1, cstep):\n",
    "                    chunk = image[rv:r, cv:c, f]\n",
    "                    dout = nx_layer.delta[i, j, f]\n",
    "                    \n",
    "                    if layer.kind == \"max\":\n",
    "                        p = np.max(chunk)\n",
    "                        #print(p, chunk)\n",
    "                        index = np.argwhere(chunk == p)[0]\n",
    "                        #print(layer.delta[rv+index[0], cv+index[1], f].shape, dout.shape)\n",
    "                        layer.delta[rv+index[0], cv+index[1], f] = dout\n",
    "                        #print(index)\n",
    "                    if layer.kind == \"min\":\n",
    "                        p = np.min(chunk)\n",
    "                        index = np.argwhere(chunk == p)[0]\n",
    "                        layer.delta[rv+index[0], cv+index[1], f] = dout\n",
    "                    if layer.kind == \"average\":\n",
    "                        p = np.mean(chunk)\n",
    "                        layer.delta[rv:r, cv:c, f] = dout\n",
    "\n",
    "\n",
    "                    j+=1\n",
    "                    cv+=cstep\n",
    "                rv+=rstep\n",
    "                i+=1\n",
    "            \n",
    "class Flatten:\n",
    "    def __init__(self, input_shape=None):\n",
    "        self.input_shape=None\n",
    "        self.output_shape = None\n",
    "        self.input_data= None\n",
    "        self.output = None\n",
    "        self.isbias = False\n",
    "        self.activation = None\n",
    "        self.parameters = 0\n",
    "        self.delta = 0\n",
    "        self.weights = 0\n",
    "        self.bias = 0\n",
    "        self.delta_weights = 0\n",
    "        self.delta_biases = 0\n",
    "        \n",
    "    def set_output_shape(self):\n",
    "        #print(self.input_shape)\n",
    "        self.output_shape = (self.input_shape[0] * self.input_shape[1] * self.input_shape[2])\n",
    "        self.weights = 0\n",
    "    def apply_activation(self, x):\n",
    "        self.input_data = x\n",
    "        #print(x.sum(axis=2))\n",
    "        self.output = np.array(self.input_data).flatten()\n",
    "        return self.output\n",
    "    def activation_dfn(self, x):\n",
    "        return x\n",
    "    def backpropagate(self, nx_layer):\n",
    "        self.error = np.dot(nx_layer.weights, nx_layer.delta)\n",
    "        self.delta = self.error * self.activation_dfn(self.out)\n",
    "        self.delta = self.delta.reshape(self.input_shape)\n",
    "\n",
    "class Conv2d():\n",
    "    def __init__(self, input_shape=None, filters=1, kernel_size = (3, 3), isbias=True, activation=None, stride=(1, 1), padding=\"zero\", kernel=None, bias=None):\n",
    "        #super().__init__()\n",
    "        self.input_shape = input_shape\n",
    "        self.filters = filters\n",
    "        self.isbias = isbias\n",
    "        self.activation = activation\n",
    "        self.stride = stride\n",
    "        self.padding = padding\n",
    "        self.p = 1 if padding != None else 0\n",
    "        self.bias = bias\n",
    "        self.kernel = kernel\n",
    "        if input_shape != None:\n",
    "            # (h, w, nc, nf)\n",
    "            self.kernel_size = (kernel_size[0], kernel_size[1], input_shape[2], filters)\n",
    "            self.output_shape = (int((input_shape[0] - kernel_size[0] + 2 * self.p) / stride[0]) + 1, \n",
    "                                int((input_shape[1] - kernel_size[1] + 2 * self.p) / stride[1]) + 1, filters)\n",
    "            self.set_variables()\n",
    "            self.out = np.zeros(self.output_shape)\n",
    "            self.dout = np.zeros((self.output_shape[0], self.output_shape[1], self.input_shape[2], self.output_shape[2]))\n",
    "        else:\n",
    "            #pass\n",
    "            # we dont know input shape yet\n",
    "            self.kernel_size = (kernel_size[0], kernel_size[1])\n",
    "            #self.set_variables()\n",
    "            \n",
    "        self.delta = 0\n",
    "        self.dfilt = 0\n",
    "        self.dbias = 0\n",
    "        self.error = 0\n",
    "        #self.out = np.zeros(self.output_shape)\n",
    "        #self.dout = np.zeros(self.output_shape)\n",
    "        self.flow = None\n",
    "        \n",
    "        # make suitable for all layers\n",
    "        #self.weights = self.kernel\n",
    "        #self.biases = self.bias\n",
    "        \n",
    "    def init_param(self, size):\n",
    "        stddev = 1/np.sqrt(np.prod(size))\n",
    "        return np.random.normal(loc=0, scale=stddev, size=size)\n",
    "    def set_variables(self):\n",
    "        \"\"\"\n",
    "        self.bias = np.random.random((self.filters, 1))\n",
    "        #self.bias = np.random.random((self.kernel_size[0], self.kernel_size[1], self.input_shape[2], self.filters)) if self.isbias == True and self.bias==None else 0\n",
    "        self.kernel = np.random.random((self.kernel_size[0], self.kernel_size[1], self.input_shape[2], self.filters)) if self.kernel == None else self.kernel\n",
    "        self.parameters = np.multiply.reduce(self.kernel_size) * 2 if self.isbias else 1        \n",
    "        self.d_final_filt = np.zeros(self.kernel.shape)\n",
    "        self.d_final_bias = np.zeros(self.bias.shape)\n",
    "        self.kernel = self.init_param(self.kernel.shape)\n",
    "        \"\"\"\n",
    "        self.weights = self.init_param(self.kernel_size)\n",
    "        self.biases = self.init_param((self.filters, 1))\n",
    "        self.parameters = np.multiply.reduce(self.kernel_size) + self.filters if self.isbias else 1\n",
    "        self.delta_weights = np.zeros(self.kernel_size)\n",
    "        self.delta_biases = np.zeros(self.biases.shape)\n",
    "        #print(self.weights.shape, self.biases.shape)\n",
    "       \n",
    "    def apply_activation(self, image):\n",
    "        #print(self.weights, self.biases)\n",
    "        for f in range(self.filters):\n",
    "            image = self.input\n",
    "            kshape = self.kernel_size\n",
    "            if kshape[0] % 2 != 1 or kshape[1] % 2 != 1:\n",
    "                raise ValueError(\"Please provide odd length of 2d kernel.\")\n",
    "\n",
    "            if type(self.stride) == int:\n",
    "                     stride = (stride, stride)\n",
    "            else:\n",
    "                stride = self.stride\n",
    "            shape = image.shape\n",
    "            if self.padding == \"zero\":\n",
    "                zeros_h = np.zeros((shape[1], shape[2])).reshape(-1, shape[1], shape[2])\n",
    "                zeros_v = np.zeros((shape[0]+2, shape[2])).reshape(shape[0]+2, -1, shape[2])\n",
    "                #print(image.shape, zeros_h.shape, zeros_v.shape)\n",
    "                #zero padding\n",
    "                padded_img = np.vstack((zeros_h, image, zeros_h)) # add rows\n",
    "                # print(padded_img)\n",
    "                padded_img = np.hstack((zeros_v, padded_img, zeros_v)) # add cols\n",
    "\n",
    "                image = padded_img\n",
    "                shape = image.shape\n",
    "\n",
    "            elif self.padding == \"same\":\n",
    "                h1 = image[0].reshape(-1, shape[1], shape[2])\n",
    "                h2 = image[-1].reshape(-1, shape[1], shape[2])\n",
    "\n",
    "\n",
    "                #zero padding\n",
    "                padded_img = np.vstack((h1, image, h2)) # add rows\n",
    "\n",
    "                v1 = padded_img[:, 0].reshape(padded_img.shape[0], -1, shape[2])\n",
    "                v2 = padded_img[:, -1].reshape(padded_img.shape[0], -1, shape[2])\n",
    "\n",
    "                padded_img = np.hstack((v1, padded_img, v2)) # add cols\n",
    "\n",
    "                image = padded_img\n",
    "                shape = image.shape\n",
    "            elif self.padding == None:\n",
    "                pass\n",
    "\n",
    "            rv = 0\n",
    "            cimg = []\n",
    "            for r in range(kshape[0], shape[0]+1, stride[0]):\n",
    "                cv = 0\n",
    "                for c in range(kshape[1], shape[1]+1, stride[1]):\n",
    "                    #if r - rv\n",
    "                    chunk = image[rv:r, cv:c]\n",
    "                    #print(chunk.shape, self.kernel[:, :, :, f].shape)\n",
    "                    # soma with each input image's chunk(can have channel) * kernel(can have channels)\n",
    "                    soma = (np.multiply(chunk, self.weights[:, :, :, f]))\n",
    "                    summa = soma.sum()+self.biases[f]\n",
    "                    #print(soma.shape)\n",
    "                    #summa = soma.mean()\n",
    "                    #print(f\"soma{soma.shape} \\n chunk{chunk.shape}, {summa.shape}\")\n",
    "                    \n",
    "                    cimg.append(summa)\n",
    "                    cv+=stride[1]\n",
    "                rv+=stride[0]\n",
    "            #print((r, c), (rv, cv))\n",
    "            cimg = np.array(cimg).reshape(int(rv/stride[0]), int(cv/stride[1]))\n",
    "            #print(cimg.shape)\n",
    "            self.out[:, :, f] = cimg\n",
    "        self.out = self.activation_fn(self.out)\n",
    "        return self.out\n",
    "        \n",
    "    \n",
    "    def activation_dfn(self, r):\n",
    "        \"\"\"\n",
    "            A method of FFL to find derivative of given activation function.\n",
    "        \"\"\"\n",
    "        \n",
    "        \n",
    "        if self.activation is None:\n",
    "            return np.ones(r.shape)\n",
    "\n",
    "        if self.activation == 'tanh':\n",
    "            return 1 - r ** 2\n",
    "        if self.activation == \"relu\":\n",
    "            r[r<0] = 0\n",
    "            return r\n",
    "        if self.activation == 'sigmoid':\n",
    "            return r * (1 - r)\n",
    "\n",
    "        if self.activation == \"softmax\":\n",
    "            soft = self.activation_fn(r)\n",
    "            diag_soft = soft * (1-soft)        \n",
    "            #s = soft.reshape(-1, 1)\n",
    "            #dsoft = np.diagflat(s) - np.dot(s, s.T)\n",
    "            #diag_soft = dsoft.diagonal()\n",
    "            # take only the diagonal of dsoft i.e i==j only\n",
    "            \"\"\"\n",
    "                soft = a / np.sum(a)\n",
    "\n",
    "                dsoft = np.diag(soft)\n",
    "                for i in range(len(x)):\n",
    "                    for j in range(len(soft)):\n",
    "                        if i == j:\n",
    "                            d = 1\n",
    "                        else:\n",
    "                            d = 0\n",
    "                        dsoft[i][j] = soft[i] * (d - soft[j])\n",
    "            \"\"\"\n",
    "            \n",
    "            return diag_soft\n",
    "        \n",
    "        return r\n",
    "    \n",
    "    def activation_fn(self, r):\n",
    "        \"\"\"\n",
    "        A method of FFL which contains the operation and defination of given activation function.\n",
    "        \"\"\"\n",
    "        \n",
    "        if self.activation == None or self.activation == \"linear\":\n",
    "            return r\n",
    "        if self.activation == \"relu\":\n",
    "            r[r<0] = 0\n",
    "            return r\n",
    "        \n",
    "        # tanh\n",
    "        if self.activation == 'tanh':\n",
    "            return np.tanh(r)\n",
    "\n",
    "        # sigmoid\n",
    "        if self.activation == 'sigmoid':\n",
    "            \n",
    "            return 1 / (1 + np.exp(-r))\n",
    "\n",
    "        if self.activation == \"softmax\":\n",
    "            # stable softmax\n",
    "            r = r - np.max(r)\n",
    "            s = np.exp(r)\n",
    "            return s / np.sum(s)\n",
    "        \n",
    "    def set_output_shape(self):\n",
    "        #print(self.input_shape, self.kernel_size, self.stride)\n",
    "        self.kernel_size = (self.kernel_size[0], self.kernel_size[1], self.input_shape[2], self.filters)\n",
    "        self.set_variables()\n",
    "        self.output_shape = (int((self.input_shape[0] - self.kernel_size[0] + 2 * self.p) / self.stride[0] + 1), \n",
    "                                int((self.input_shape[1] - self.kernel_size[1] + 2 * self.p) / self.stride[1] + 1), self.filters)\n",
    "        self.out = np.zeros(self.output_shape)\n",
    "        self.dout = np.zeros((self.output_shape[0], self.output_shape[1], self.input_shape[2], self.output_shape[2]))\n",
    "    \n",
    "    def backpropagate(self, nx_layer):\n",
    "        layer = self\n",
    "        #layer.dfilt = np.zeros(layer.kernel.shape)\n",
    "        #layer.delta = np.zeros(layer.input_shape)\n",
    "        #layer.dbias = np.zeros(layer.bias.shape)\n",
    "        layer.delta = np.zeros((layer.input_shape[0], layer.input_shape[1], layer.input_shape[2]))\n",
    "        \n",
    "        image = layer.input\n",
    "        #print(layer.name, layer.delta.shape)\n",
    "        #print(f\"nx lyr's delshape\", nx_layer.delta.shape, layer.input_shape)\n",
    "        #print(f\"{nx_layer.delta.shape, layer.kernel.shape}\")\n",
    "         \n",
    "        for f in range(layer.filters):\n",
    "            kshape = layer.kernel_size\n",
    "            shape = layer.input_shape\n",
    "            stride = layer.stride\n",
    "            rv = 0\n",
    "            i = 0\n",
    "            #print(layer.kernel.shape)\n",
    "            for r in range(kshape[0], shape[0]+1, stride[0]):\n",
    "                cv = 0\n",
    "                j = 0\n",
    "                for c in range(kshape[1], shape[1]+1, stride[1]):\n",
    "                    chunk = image[rv:r, cv:c]\n",
    "                    #print(layer.name, chunk.shape, layer.dfilt.shape, nx_layer.delta.shape)\n",
    "                    #\"\"\"\n",
    "                    #original\n",
    "                    #print(layer.delta_weights.shape)\n",
    "                    layer.delta_weights[:, :, :, f] += chunk * nx_layer.delta[i, j, f]\n",
    "                    \n",
    "                    #layer.delta[rv:r, cv:c] += nx_layer.delta[i, j] * layer.dfilt[f]\n",
    "                    #print(\"upd\", layer.delta[rv:r, cv:c, f].shape, nx_layer.delta[i, j, f].shape, layer.kernel[:, :, :, f].shape)\n",
    "                    #print(layer.name, layer.kernel.shape)\n",
    "                    layer.delta[rv:r, cv:c, :] += nx_layer.delta[i, j, f] * layer.weights[:, :, :, f]\n",
    "                    #chunk = int((np.multiply(chunk, kernel)+bias).sum())\n",
    "                    j+=1\n",
    "                    cv+=stride[1]\n",
    "                rv+=stride[0]\n",
    "                i+=1\n",
    "            #layer.kernel[f] += dfilt\n",
    "            layer.delta_biases[f] = np.sum(nx_layer.delta[:, :, f])\n",
    "        layer.delta = layer.activation_dfn(layer.delta)\n",
    "        \n",
    "\n",
    "\"\"\"New CNN.\"\"\"        \n",
    "class CNN():\n",
    "    def __init__(self):\n",
    "        self.layers = []\n",
    "        self.info_df = {}\n",
    "        self.column = [\"LName\", \"Input Shape\", \"Output Shape\", \"Activation\", \"Bias\"]\n",
    "        self.parameters = []\n",
    "        self.optimizer = \"\"\n",
    "        self.loss = \"mse\"\n",
    "        #self.all_loss = {}\n",
    "        self.lr = 0.01\n",
    "        self.mr = 0.0001\n",
    "        self.metrics = []\n",
    "        self.av_optimizers = [\"sgd\", \"momentum\", \"adam\"]\n",
    "        self.av_metrics = [\"mse\", \"accuracy\", \"cse\"]\n",
    "        self.av_loss = [\"mse\", \"cse\"]\n",
    "        self.iscompiled = False\n",
    "        self.model_dict = None\n",
    "        self.out = []\n",
    "        self.eps = 1e-15\n",
    "        #self.non_trainable=[\"Flatten\"]\n",
    "        self.train_loss = {}\n",
    "        self.val_loss = {}\n",
    "        self.train_acc = {}\n",
    "        self.val_acc = {}\n",
    "    \n",
    "    def add(self, layer):\n",
    "        if(len(self.layers) > 0):\n",
    "            prev_layer = self.layers[-1]\n",
    "            if prev_layer.name != \"Input Layer\":\n",
    "                prev_layer.name = f\"{type(prev_layer).__name__}{len(self.layers) - 1}\" \n",
    "            \n",
    "            if layer.input_shape == None:\n",
    "                if type(layer).__name__ == \"Flatten\":\n",
    "                        ops = prev_layer.output_shape[:]\n",
    "                        if type(prev_layer).__name__ == \"Pool2d\":\n",
    "                            ops = prev_layer.output_shape[:]\n",
    "                elif type(layer).__name__ == \"Conv2d\":\n",
    "                    ops = prev_layer.output_shape[:]\n",
    "                    if type(prev_layer).__name__ == \"Pool2d\":\n",
    "                        ops = prev_layer.output_shape\n",
    "                elif type(layer).__name__ == \"Pool2d\":\n",
    "                    ops = prev_layer.output_shape[:]\n",
    "                    if type(prev_layer).__name__ == \"Pool2d\":\n",
    "                        ops = prev_layer.output_shape[:]\n",
    "                    #print(prev_layer.output_shape)\n",
    "                    \n",
    "                else:\n",
    "                    ops = prev_layer.output_shape\n",
    "                layer.input_shape = ops\n",
    "#                 print(prev_layer.output_shape)\n",
    "                #print(type(layer).__name__, layer.input_shape)\n",
    "                layer.set_output_shape()\n",
    "            layer.name = f\"Out Layer({type(layer).__name__})\"\n",
    "            #if prev_layer.neurons != layer.n_input:\n",
    "             #   raise ValueError(f\"This layer '{layer.name}' must have neurons={prev_layer.neurons} because '{prev_layer.name}' has output of {prev_layer.neurons}.\")\n",
    "        else:\n",
    "            layer.name = \"Input Layer\"\n",
    "        if type(layer).__name__ == \"Conv2d\":\n",
    "            if(layer.output_shape[0] <= 0 or layer.output_shape[1] <= 0):\n",
    "                raise ValueError(f\"The output shape became invalid [i.e. {layer.output_shape}]. Reduce filter size or increase image size.\")\n",
    "        self.layers.append(layer)\n",
    "        self.parameters.append(layer.parameters)\n",
    "        \n",
    "    def summary(self):\n",
    "        lname = []\n",
    "        linput = []\n",
    "        loutput = []\n",
    "        lactivation = []\n",
    "        lisbias = []\n",
    "        lparam = []\n",
    "        for layer in self.layers:\n",
    "            lname.append(layer.name)\n",
    "            linput.append(layer.input_shape)\n",
    "            loutput.append(layer.output_shape)########################\n",
    "            lactivation.append(layer.activation)\n",
    "            lisbias.append(layer.isbias)\n",
    "            lparam.append(layer.parameters)\n",
    "        model_dict = {\"Layer Name\": lname, \"Input\": linput, \"Output Shape\": loutput,\n",
    "                      \"Activation\": lactivation, \"Bias\": lisbias, \"Parameters\": lparam}    \n",
    "        model_df = pd.DataFrame(model_dict).set_index(\"Layer Name\")\n",
    "        print(model_df)\n",
    "        print(f\"Total Parameters: {sum(lparam)}\")\n",
    "    \n",
    "    def apply_loss(self, y, out):\n",
    "        if self.loss == \"mse\":\n",
    "            #print(output_shape)\n",
    "            loss = y - out\n",
    "            mse = np.mean(np.square(loss), axis=-1)\n",
    "#             self.all_loss.append(mse)          \n",
    "            return loss, mse\n",
    "\n",
    "        if self.loss == 'cse':\n",
    "            \"\"\" Requires out to be probability values. \"\"\"     \n",
    "            if len(out) == len(y) == 1:\n",
    "                #print(\"Using Binary CSE.\")\n",
    "                #y += self.eps\n",
    "                #out += self.eps\n",
    "                cse = -(y * np.log(out) + (1 - y) * np.log(1 - out))\n",
    "                loss = -(y / out - (1 - y) / (1 - out))\n",
    "                #cse = np.mean(abs(cse))\n",
    "            else:\n",
    "                #print(\"Using Categorical CSE.\")\n",
    "                if self.layers[-1].activation == \"softmax\":\n",
    "                    # if o/p layer's fxn is softmax then loss is y - out\n",
    "                    # check the derivation of softmax and crossentropy with derivative\n",
    "                    loss = y - out\n",
    "                    loss = loss / self.layers[-1].activation_dfn(out)\n",
    "                else:\n",
    "                    y = np.float64(y)\n",
    "                    #y += self.eps\n",
    "                    out += self.eps\n",
    "                    #cse =  -np.sum(y * (np.log(out)))\n",
    "                    \n",
    "                    loss = -(np.nan_to_num(y / out, posinf=0, neginf=0) - np.nan_to_num((1 - y) / (1 - out), posinf=0, neginf=0))\n",
    "                \n",
    "            \n",
    "                cse = -np.sum(y * np.nan_to_num(np.log(out), posinf=0, neginf=0) + (1 - y) * np.nan_to_num(np.log(1 - out), posinf=0, neginf=0))\n",
    "            return loss, cse\n",
    "    \n",
    "    def compile_model(self, lr=0.01, mr = 0.001, opt = \"sgd\", loss = \"mse\", metrics=['mse']):\n",
    "        if opt not in self.av_optimizers:\n",
    "            raise ValueError(f\"Optimizer is not understood, use one of {self.av_optimizers}.\")\n",
    "        \n",
    "        for m in metrics:\n",
    "            if m not in self.av_metrics:\n",
    "                raise ValueError(f\"Metrics is not understood, use one of {self.av_metrics}.\")\n",
    "        \n",
    "        if loss not in self.av_loss:\n",
    "            raise ValueError(f\"Loss function is not understood, use one of {self.av_loss}.\")\n",
    "        \n",
    "        self.optimizer = opt\n",
    "        self.loss = loss\n",
    "        self.lr = lr\n",
    "        self.mr = mr\n",
    "        self.metrics = metrics\n",
    "        self.iscompiled = True\n",
    "        self.optimizer = Optimizer(layers=self.layers, name=opt, learning_rate=lr, mr=mr)\n",
    "        self.optimizer = self.optimizer.opt_dict[opt]\n",
    "        \"\"\"\n",
    "            if opt == \"sgd\":\n",
    "                self.optimizer = self.SGD\n",
    "            elif opt == \"adam\":\n",
    "                self.optimizer = self.ADAM\n",
    "                alpha = self.lr\n",
    "                beta1 = 0.9\n",
    "                beta2 = 0.999\n",
    "                epsilon = 1e-8\n",
    "                for l in self.layers:\n",
    "                    if type(l).__name__ == \"FFL\":\n",
    "                        l.adam_param = []\n",
    "                        l.dprev = [0]\n",
    "                        # w, a, b1, b2, eps, m, v, t\n",
    "                        l.adam_param = [[l.weights, alpha, beta1, beta2, epsilon, 0, 0, 0]]\n",
    "                        if l.isbias:\n",
    "                            l.adam_param.append([l.bias, alpha, beta1, beta2, epsilon, 0, 0, 0])\n",
    "                            l.dprev.append(0)\n",
    "                    if type(l).__name__ == \"Conv2d\":\n",
    "                        l.adam_param = []\n",
    "                        l.dprev = [0]\n",
    "                        l.adam_param = [[l.kernel, alpha, beta1, beta2, epsilon, 0, 0, 0]]\n",
    "                        if l.isbias:\n",
    "                            l.adam_param.append([l.bias, alpha, beta1, beta2, epsilon, 0, 0, 0])\n",
    "                            l.dprev.append(0)\n",
    "        \"\"\"\n",
    "    def SGD(self, update):\n",
    "        for i in range(len(self.layers)):    \n",
    "            layer = self.layers[i]\n",
    "            \n",
    "            if type(layer).__name__ == \"FFL\":\n",
    "                input_to_use = np.atleast_2d(layer.input)\n",
    "                dw = layer.delta * input_to_use.T \n",
    "                db =  layer.delta \n",
    "                if update:\n",
    "                    layer.weights += (layer.dw * self.mr / self.batch_size + dw * self.lr)\n",
    "                    layer.bias += (layer.db * self.mr / self.batch_size +  db * self.lr)\n",
    "                    layer.dw = 0\n",
    "                    layer.db = 0\n",
    "                else:\n",
    "                    layer.dw += dw \n",
    "                    layer.db += db \n",
    "            if type(layer).__name__ == \"Conv2d\":\n",
    "                if update != True:\n",
    "                    layer.kernel += (layer.d_final_filt * self.mr/ self.batch_size + layer.dfilt * self.lr)\n",
    "                    layer.bias += (layer.d_final_bias * self.mr/ self.batch_size + layer.dbias * self.lr)\n",
    "                    layer.d_final_filt = 0\n",
    "                    layer.d_final_bias = 0\n",
    "                else:\n",
    "                    layer.d_final_filt += layer.dfilt \n",
    "                    layer.d_final_bias += layer.dbias \n",
    "                \n",
    "    def ADAM(self, update):\n",
    "        \n",
    "        for i in range(len(self.layers)):\n",
    "            layer = self.layers[i]       \n",
    "            if type(layer).__name__ == \"FFL\":\n",
    "                input_to_use = np.atleast_2d(layer.input)\n",
    "                dw = layer.delta * input_to_use.T \n",
    "                db =  layer.delta \n",
    "                dparam = [dw, db]\n",
    "                nparam = []\n",
    "                l = layer\n",
    "                for i, grad in enumerate(dparam):\n",
    "                    #print(i)\n",
    "                    theta = l.adam_param[i][0]\n",
    "                    alpha = l.adam_param[i][1]\n",
    "                    beta1 = l.adam_param[i][2]\n",
    "                    beta2 = l.adam_param[i][3]\n",
    "                    epsilon = l.adam_param[i][4]\n",
    "                    m = l.adam_param[i][5]\n",
    "                    v = l.adam_param[i][6]\n",
    "                    t = l.adam_param[i][7] + 1\n",
    "\n",
    "                    m = beta1 * m + (1-beta1) * grad # moving avg of grad\n",
    "                    v = beta2 * v + (1-beta2) * np.power(grad, 2) #\n",
    "                    m_cap = m / (1 - np.power(beta1, t))\n",
    "                    v_cap = v / (1 - np.power(beta2, t))\n",
    "                    theta += (self.mr * layer.dprev[i]) + np.nan_to_num(alpha * (m_cap / (np.sqrt(v_cap) + epsilon)))\n",
    "                    layer.dprev[i] = np.nan_to_num((m_cap / (np.sqrt(v_cap) + epsilon))) * 0\n",
    "                    l.adam_param[i] = [theta, alpha, beta1, beta2, epsilon, m, v, t]\n",
    "                    nparam.append(theta)\n",
    "                if len(nparam) == 2:\n",
    "                    layer.weights = nparam[0]\n",
    "                    layer.bias = nparam[1]\n",
    "                else:\n",
    "                    layer.weights = nparam[0]\n",
    "            if type(layer).__name__ == \"Conv2d\":\n",
    "                #layer.kernel += (layer.d_final_filt * self.mr/ self.batch_size + layer.dfilt * self.lr)\n",
    "                #layer.bias += (layer.d_final_bias * self.mr/ self.batch_size + layer.dbias * self.lr)\n",
    "                dw = layer.dfilt\n",
    "                db =  layer.dbias \n",
    "                dparam = [dw, db]\n",
    "                nparam = []\n",
    "                l = layer\n",
    "                for i, grad in enumerate(dparam):\n",
    "                    #print(i)\n",
    "                    theta = l.adam_param[i][0]\n",
    "                    alpha = l.adam_param[i][1]\n",
    "                    beta1 = l.adam_param[i][2]\n",
    "                    beta2 = l.adam_param[i][3]\n",
    "                    epsilon = l.adam_param[i][4]\n",
    "                    m = l.adam_param[i][5]\n",
    "                    v = l.adam_param[i][6]\n",
    "                    t = l.adam_param[i][7] + 1\n",
    "\n",
    "                    m = beta1 * m + (1-beta1) * grad # moving avg of grad\n",
    "                    v = beta2 * v + (1-beta2) * np.power(grad, 2) #\n",
    "                    m_cap = m / (1 - np.power(beta1, t))\n",
    "                    v_cap = v / (1 - np.power(beta2, t))\n",
    "                    theta += (self.mr * layer.dprev[i]) + np.nan_to_num(alpha * (m_cap / (np.sqrt(v_cap) + epsilon)))\n",
    "                    layer.dprev[i] = np.nan_to_num((m_cap / (np.sqrt(v_cap) + epsilon))) * 0\n",
    "                    l.adam_param[i] = [theta, alpha, beta1, beta2, epsilon, m, v, t]\n",
    "                    nparam.append(theta)\n",
    "                if len(nparam) == 2:\n",
    "                    layer.kernel = nparam[0]\n",
    "                    layer.bias = nparam[1]\n",
    "                else:\n",
    "                    layer.kernel = nparam[0]\n",
    "    def feedforward(self, x, train=True):\n",
    "        if train:\n",
    "            for l in self.layers:\n",
    "                l.input = x            \n",
    "                x = np.nan_to_num(l.apply_activation(x))\n",
    "                #print(l.name, x.shape)\n",
    "                l.out = x\n",
    "\n",
    "            return x\n",
    "        else:\n",
    "            for l in self.layers:\n",
    "                l.input = x \n",
    "                if type(l).__name__ == \"Dropout\":\n",
    "                    x = np.nan_to_num(l.apply_activation(x, train=train))\n",
    "                else:           \n",
    "                    x = np.nan_to_num(l.apply_activation(x))\n",
    "                l.out = x\n",
    "\n",
    "            return x\n",
    "\n",
    "    def backpropagate_conv2d(self, layer, nx_layer):\n",
    "        layer.dfilt = np.zeros(layer.kernel.shape)\n",
    "        #layer.delta = np.zeros(layer.input_shape)\n",
    "        layer.dbias = np.zeros(layer.kernel.shape)\n",
    "        layer.delta = np.zeros((layer.input_shape[0], layer.input_shape[1], layer.filters))\n",
    "        \n",
    "        image = layer.input\n",
    "        #print(layer.name)\n",
    "        #print(f\"nx lyr's delshape\", nx_layer.delta.shape, layer.input_shape)\n",
    "        #print(f\"{nx_layer.delta.shape, layer.kernel.shape}\")\n",
    "         \n",
    "        for f in range(layer.filters):\n",
    "            kshape = layer.kernel.shape\n",
    "            shape = layer.input_shape\n",
    "            stride = layer.stride\n",
    "            rv = 0\n",
    "            i = 0\n",
    "            #print(layer.kernel.shape)\n",
    "            for r in range(kshape[0], shape[0]+1, stride[0]):\n",
    "                cv = 0\n",
    "                j = 0\n",
    "                for c in range(kshape[1], shape[1]+1, stride[1]):\n",
    "                    chunk = image[rv:r, cv:c]\n",
    "                    #print(layer.name, chunk.shape, nx_layer.delta[i, j, f], layer.dfilt[:, :, :, f].shape)\n",
    "                    #\"\"\"\n",
    "                    #original\n",
    "                    layer.dfilt[:, :, :, f] += chunk * nx_layer.delta[i, j, f]\n",
    "                    \n",
    "                    #print(f\"dsh {layer.kernel[:, :, :, f].shape} nld{nx_layer.delta[i, j, f]}\")\n",
    "                    #layer.delta[rv:r, cv:c] += nx_layer.delta[i, j] * layer.dfilt[f]\n",
    "                    #print(\"upd\", layer.delta[rv:r, cv:c, f].shape, layer.kernel[:, :, :, f].shape)\n",
    "                    layer.delta[rv:r, cv:c, f] += nx_layer.delta[i, j, f] * layer.kernel[:, :, :, f].sum(axis = 2)\n",
    "                    \n",
    "                    \"\"\"\n",
    "                    e = nx_layer.delta[i, j, f] * filt\n",
    "                    d = e * nx_layer.input[i, j]\n",
    "                    df = e * chunk\n",
    "                    layer.dfilt[f] += df\n",
    "                    layer.delta[rv:r, cv:c] += d\n",
    "                    \"\"\"\n",
    "                    #chunk = int((np.multiply(chunk, kernel)+bias).sum())\n",
    "                    j+=1\n",
    "                    cv+=stride[1]\n",
    "                rv+=stride[0]\n",
    "                i+=1\n",
    "            #layer.kernel[f] += dfilt\n",
    "            layer.dbias[:, :, :, f] = np.sum(layer.delta[:, :, f])\n",
    "            #print(layer.dbias[f].shape, layer.bias[f].shape)\n",
    "            \n",
    "            #cimg = np.array(cimg, dtype=np.uint8).reshape(int(rv/stride[0]), int(cv/stride[1])) \n",
    "                \n",
    "    def backpropagate_pool2d(self, layer, nx_layer):\n",
    "        \"\"\"\n",
    "            Gradients are passed through index of largest value .\n",
    "        \"\"\"\n",
    "        stride = layer.stride\n",
    "        kshape = layer.kernel_size\n",
    "        image = layer.input\n",
    "        shape = image.shape\n",
    "        layer.delta = np.zeros(shape)\n",
    "        #self.input_shape = shape\n",
    "        #self.set_output_shape()\n",
    "        #print(self.output_shape)\n",
    "        \n",
    "        cimg = []\n",
    "        rstep = stride[0]+kshape[0]-1\n",
    "        cstep = stride[1]+kshape[1]-1\n",
    "        \n",
    "        for f in range(shape[2]):\n",
    "            i = 0\n",
    "            rv = 0\n",
    "            for r in range(kshape[0], shape[0]+1, rstep):\n",
    "                cv = 0\n",
    "                j = 0\n",
    "                for c in range(kshape[1], shape[1]+1, cstep):\n",
    "                    chunk = image[rv:r, cv:c, f]\n",
    "                    dout = nx_layer.delta[i, j, f]\n",
    "                    \n",
    "                    if layer.kind == \"max\":\n",
    "                        p = np.max(chunk)\n",
    "                        #print(p, chunk)\n",
    "                        index = np.argwhere(chunk == p)[0]\n",
    "                        #print(layer.delta[rv+index[0], cv+index[1], f].shape, dout.shape)\n",
    "                        layer.delta[rv+index[0], cv+index[1], f] = dout\n",
    "                        #print(index)\n",
    "                    if layer.kind == \"min\":\n",
    "                        p = np.min(chunk)\n",
    "                        index = np.argwhere(chunk == p)[0]\n",
    "                        layer.delta[rv+index[0], cv+index[1], f] = dout\n",
    "                    if layer.kind == \"average\":\n",
    "                        p = np.mean(chunk)\n",
    "                        layer.delta[rv:r, cv:c, f] = dout\n",
    "\n",
    "\n",
    "                    j+=1\n",
    "                    cv+=cstep\n",
    "                rv+=rstep\n",
    "                i+=1\n",
    "            \n",
    "        #print(layer.name, layer.delta.shape, layer.input_shape)\n",
    "#         layer.delta = (layer.input/(layer.input.sum(axis=2).reshape(layer.input_shape[0], layer.input_shape[1], 1)) * layer.delta.reshape(layer.input_shape[0], layer.input_shape[1], 1))\n",
    "                            \n",
    "    def backpropagate(self, loss, update):\n",
    "        \n",
    "        # if it is output layer\n",
    "        for i in reversed(range(len(self.layers))):\n",
    "            layer = self.layers[i]\n",
    "            if layer == self.layers[-1]:\n",
    "                if (type(layer).__name__ == \"FFL\"):\n",
    "                    layer.error = loss\n",
    "                    layer.delta = layer.error * layer.activation_dfn(layer.out)\n",
    "                    layer.delta_weights += layer.delta * np.atleast_2d(layer.input).T\n",
    "                    layer.delta_biases += layer.delta\n",
    "            else:\n",
    "                nx_layer = self.layers[i+1]\n",
    "                \"\"\"\n",
    "                    \n",
    "                    if (type(layer).__name__ == 'FFL' or type(nx_layer).__name__ == 'FFL'):\n",
    "                        layer.error = np.dot(nx_layer.weights, nx_layer.delta)\n",
    "                        layer.delta = layer.error * layer.activation_dfn(layer.out)\n",
    "\n",
    "                        if type(layer).__name__ == \"Flatten\":\n",
    "                            layer.delta = layer.delta.reshape(layer.input_shape)\n",
    "\n",
    "\n",
    "                    if type(layer).__name__ == \"Dropout\":\n",
    "                        if type(nx_layer).__name__ != \"Conv2d\":\n",
    "                            layer.error = np.dot(nx_layer.weights, nx_layer.delta)\n",
    "                            layer.delta = layer.error * layer.activation_dfn(layer.out)\n",
    "                        else:\n",
    "                            layer.delta = nx_layer.delta\n",
    "                    if type(layer).__name__ == \"Conv2d\":\n",
    "                        self.backpropagate_conv2d(layer=layer, nx_layer=nx_layer)\n",
    "                    if type(layer).__name__ == \"Pool2d\":\n",
    "                        self.backpropagate_pool2d(layer=layer, nx_layer=nx_layer)\n",
    "                \"\"\"\n",
    "                #print(layer.name)\n",
    "                layer.backpropagate(nx_layer)\n",
    "            if update:\n",
    "                layer.delta_weights /= self.batch_size\n",
    "                layer.delta_biases /= self.batch_size\n",
    "        if update: \n",
    "            self.optimizer(self.layers)\n",
    "            self.zerograd()\n",
    "    def zerograd(self):\n",
    "        for l in self.layers:\n",
    "            try:\n",
    "                l.delta_weights=np.zeros(l.delta_weights.shape)\n",
    "                l.delta_biases = np.zeros(l.delta_biases.shape)\n",
    "            except:\n",
    "                pass\n",
    "        #self.optimizer(update)\n",
    "\n",
    "    def predict(self, X):\n",
    "        out = []\n",
    "        if X.shape != self.layers[0].input_shape:\n",
    "            for x in X:\n",
    "                out.append(self.feedforward(x, train=False))\n",
    "            \n",
    "        else:\n",
    "            out.append(self.feedforward(X, train = False))\n",
    "        return np.array(out)\n",
    "    def accuracy_score(self, y, yt):\n",
    "        pass\n",
    "    def check_trainnable(self, X, Y):\n",
    "        #print(X[0].shape, self.layers[0].input_shape)\n",
    "        if self.iscompiled == False:\n",
    "            raise ValueError(\"Model is not compiled.\")\n",
    "        if len(X) != len(Y):\n",
    "            raise ValueError(\"Length of training input and label is not equal.\")\n",
    "        if X[0].shape != self.layers[0].input_shape:\n",
    "            layer = self.layers[0]\n",
    "            raise ValueError(f\"'{layer.name}' expects input of {layer.input_shape} while {X[0].shape[0]} is given.\")\n",
    "        if Y.shape[-1] != self.layers[-1].neurons:\n",
    "            op_layer = self.layers[-1]\n",
    "            raise ValueError(f\"'{op_layer.name}' expects input of {op_layer.neurons} while {Y.shape[-1]} is given.\")  \n",
    "        \n",
    "    def train(self, X, Y, epochs, show_every=1, batch_size = 32, shuffle=True, val_split=0.1, val_x=None, val_y=None):     \n",
    "        self.check_trainnable(X, Y)\n",
    "        self.batch_size = batch_size\n",
    "        t1 = time.time()\n",
    "        curr_ind = np.arange(0, len(X), dtype=np.int32)\n",
    "        if shuffle: \n",
    "            np.random.shuffle(curr_ind)\n",
    "           \n",
    "        if type(val_x) != type(None) and type(val_y) != type(None):\n",
    "            self.check_trainnable(val_x, val_y)\n",
    "            print(\"\\nValidation data found.\\n\")\n",
    "        else:\n",
    "            val_ex = int(len(X) * val_split)\n",
    "            val_exs = []\n",
    "            while len(val_exs) != val_ex:\n",
    "                rand_ind = np.random.randint(0, len(X))\n",
    "                if rand_ind not in val_exs:\n",
    "                    val_exs.append(rand_ind)\n",
    "            val_ex = np.array(val_exs)\n",
    "            val_x, val_y = X[val_ex], Y[val_ex]\n",
    "            curr_ind = np.array([v for v in curr_ind if v not in val_ex])\n",
    "                 \n",
    "        print(f\"\\nTotal {len(X)} samples.\\nTraining samples: {len(curr_ind)} Validation samples: {len(val_x)}.\")        \n",
    "        out_activation = self.layers[-1].activation\n",
    "        batches = []\n",
    "        len_batch = int(len(curr_ind)/batch_size) \n",
    "        if len(curr_ind)%batch_size != 0:\n",
    "            len_batch+=1\n",
    "        batches = np.array_split(curr_ind, len_batch)\n",
    "        \n",
    "        print(f\"Total {len_batch} batches, most batch has {batch_size} samples.\\n\")\n",
    "        for e in range(epochs):            \n",
    "            err = []\n",
    "            for batch in batches:\n",
    "                a = [] \n",
    "                curr_x, curr_y = X[batch], Y[batch]\n",
    "                b = 0\n",
    "                batch_loss = 0\n",
    "                for x, y in zip(curr_x, curr_y):\n",
    "                    out = self.feedforward(x)\n",
    "                    loss, error = self.apply_loss(y, out)\n",
    "                    #loss = loss.mean(axis=0)\n",
    "                    batch_loss += loss\n",
    "                    err.append(error)\n",
    "                    update = False\n",
    "                    \n",
    "                    if b == batch_size-1:\n",
    "                        update = True\n",
    "                        loss = batch_loss/batch_size\n",
    "                    self.backpropagate(loss, update)\n",
    "                    b+=1\n",
    "                    \n",
    "              \n",
    "            if e % show_every == 0:      \n",
    "                train_out = self.predict(X[curr_ind])\n",
    "                train_loss, train_error = self.apply_loss(Y[curr_ind], train_out)\n",
    "                \n",
    "                val_out = self.predict(val_x)\n",
    "                val_loss, val_error = self.apply_loss(val_y, val_out)\n",
    "                \n",
    "                if out_activation == \"softmax\":\n",
    "                    train_acc = train_out.argmax(axis=1) == Y[curr_ind].argmax(axis=1)\n",
    "                    val_acc = val_out.argmax(axis=1) == val_y.argmax(axis=1)\n",
    "                elif out_activation == \"sigmoid\":\n",
    "                    train_acc = train_out > 0.7\n",
    "                    val_acc = val_out > 0.7\n",
    "                    #pred = pred == Y\n",
    "                elif out_activation == None:\n",
    "                    train_acc = abs(Y[curr_ind]-train_out) < 0.000001\n",
    "                    val_acc = abs(Y[val_ex]-val_out) < 0.000001\n",
    "                    \n",
    "                self.train_loss[e] = round(train_error.mean(), 4)\n",
    "                self.train_acc[e] = round(train_acc.mean() * 100, 4)\n",
    "                \n",
    "                self.val_loss[e] = round(val_error.mean(), 4)\n",
    "                self.val_acc[e] = round(val_acc.mean()*100, 4)\n",
    "                print(f\"Epoch: {e}:\")\n",
    "                print(f\"Time: {round(time.time() - t1, 3)}sec\")\n",
    "                print(f\"Train Loss: {round(train_error.mean(), 4)} Train Accuracy: {round(train_acc.mean() * 100, 4)}%\")\n",
    "                print(f'Val Loss: {(round(val_error.mean(), 4))} Val Accuracy: {round(val_acc.mean() * 100, 4)}% \\n')\n",
    "                \n",
    "                t1 = time.time()\n",
    "\n",
    "                \n",
    "from keras.datasets import mnist\n",
    "(x_train, y_train), (x_test, y_test)  = mnist.load_data()\n",
    "x = x_train.reshape(-1, 28 * 28) \n",
    "x = (x-x.mean(axis=1).reshape(-1, 1))/x.std(axis=1).reshape(-1, 1)\n",
    "x = x.reshape(-1, 28, 28, 1) \n",
    "y = pd.get_dummies(y_train).to_numpy()\n",
    "\n",
    "xt = x_test.reshape(-1, 28 * 28) \n",
    "xt = (xt-xt.mean(axis=1).reshape(-1, 1))/xt.std(axis=1).reshape(-1, 1)\n",
    "xt = xt.reshape(-1, 28, 28, 1) \n",
    "yt = pd.get_dummies(y_test).to_numpy()\n",
    "\n",
    "# use devanagari\n",
    "# test = pd.read_csv(\"Digits/test.csv\")\n",
    "# train = pd.read_csv(\"Digits/train.csv\")\n",
    "# trainn = train.to_numpy()\n",
    "# x_train = trainn[:, 1:].reshape(-1, 32, 32, 1) / 255\n",
    "# y_train = trainn[:, 0]\n",
    "# y_train = pd.get_dummies(y_train).to_numpy()\n",
    "\n",
    "# testn = test.to_numpy()\n",
    "# x_test = testn[:, 1:].reshape(-1, 32, 32, 1) / 255\n",
    "# y_test = testn[:, 0]\n",
    "# y_test = pd.get_dummies(y_test).to_numpy() \n",
    "    \n",
    "m = CNN()\n",
    "m.add(Conv2d(input_shape = (28, 28, 1), filters = 4, padding=None, kernel_size=(3, 3), activation=\"relu\"))\n",
    "#m.add(Pool2d(kernel_size=(2, 2)))\n",
    "m.add(Conv2d(filters=8, kernel_size=(3, 3), padding=None, activation=\"relu\"))\n",
    "m.add(Dropout(0.2))\n",
    "m.add(Upsample())\n",
    "m.add(Flatten())\n",
    "m.add(FFL(neurons = 10, activation='softmax'))\n",
    "m.compile_model(lr=0.01, opt=\"adam\", loss=\"cse\")\n",
    "m.summary()\n",
    "m.train(x[:100], y[:100], epochs=100, batch_size=32, val_x=xt[:5], val_y=yt[:5]) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-11T03:35:47.337118Z",
     "start_time": "2020-06-11T03:35:47.005307Z"
    },
    "code_folding": [
     0,
     38
    ]
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model Loaded...\n",
      "                      Input Output Shape Activation   Bias  Parameters\n",
      "Layer Name                                                            \n",
      "Input Layer     (28, 28, 1)  (26, 26, 4)       relu   True          40\n",
      "Pool2d1         (26, 26, 4)  (13, 13, 4)       None  False           0\n",
      "Conv2d2         (13, 13, 4)  (11, 11, 8)       relu   True         296\n",
      "Dropout3        (11, 11, 8)  (11, 11, 8)       None  False           0\n",
      "Flatten4        (11, 11, 8)          968       None  False           0\n",
      "Out Layer(FFL)          968           10    softmax   True        9690\n",
      "Total Parameters: 10026\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[False, False, False, False, False, False, False, False, False,\n",
       "        False]])"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    " def save_model(self, path=\"model.json\"):\n",
    "        \"\"\"\n",
    "            path:- where to save a model including filename\n",
    "            saves Json file on given path.\n",
    "        \"\"\"\n",
    "\n",
    "        dict_model = {\"model\":str(type(self).__name__)}\n",
    "        to_save = [\"name\", \"isbias\", \"neurons\", \"input_shape\", \"output_shape\", \n",
    "                   \"weights\", \"biases\", \"activation\", \"parameters\", \"filters\",\n",
    "                  \"kernel_size\", \"padding\", \"prob\", \"stride\", \"kind\"]\n",
    "        for l in self.layers:\n",
    "            current_layer = vars(l)\n",
    "            #print(current_layer)\n",
    "            values = {\"type\":str(type(l).__name__)}\n",
    "            for key, value in current_layer.items():\n",
    "                if key in to_save:\n",
    "                    if key in [\"weights\", \"biases\"]:\n",
    "                        try:\n",
    "                            value = value.tolist()\n",
    "                        except:\n",
    "                            value = float(value)\n",
    "                        #print(value)\n",
    "                    if type(value)== np.int32:\n",
    "                        value = float(value)\n",
    "                    if key == \"input_shape\" or key == \"output_shape\":\n",
    "                        try:\n",
    "                            value = tuple(value)\n",
    "                        except:\n",
    "                            pass\n",
    "                    values[key] = value\n",
    "            # not all values will be set because many layers can be FFL and this key will be replaced    \n",
    "            dict_model[l.name] = values\n",
    "        json_dict = json.dumps(dict_model)    \n",
    "        with open(path, mode=\"w\") as f:\n",
    "            f.write(json_dict)\n",
    "            #print(json_dict)\n",
    "        print(\"\\nModel Saved.\")\n",
    "#save_model(m)\n",
    "def load_model(path=\"model.json\"):\n",
    "    \"\"\"\n",
    "        path:- path of model file including filename        \n",
    "        returns:- a model\n",
    "    \"\"\"\n",
    "    \n",
    "    models = {\"CNN\": CNN}\n",
    "    #layers = {\"FFL\": FFL}\n",
    "    layers = {\"FFL\": FFL, \"Conv2d\":Conv2d, \"Dropout\":Dropout, \"Flatten\": Flatten, \"Pool2d\":Pool2d}\n",
    "    with open(path, \"r\") as f:\n",
    "        dict_model = json.load(f)\n",
    "        model = dict_model[\"model\"]\n",
    "        #exec(\"model=models[model]\")\n",
    "        model = models[model]()\n",
    "        #exec(\"model=model()\")\n",
    "        for layer, params in dict_model.items():\n",
    "            if layer != \"model\":\n",
    "                # create a layer obj\n",
    "                lyr_type = layers[params[\"type\"]]\n",
    "                #print(layers[params[\"type\"]], Conv2d)\n",
    "                \n",
    "                ###### create models here.                \n",
    "                if lyr_type == FFL:                                        \n",
    "                    lyr.neurons = params[\"neurons\"]\n",
    "                    lyr = layers[params[\"type\"]](neurons=params[\"neurons\"])\n",
    "                \n",
    "                if lyr_type == Conv2d:\n",
    "                    lyr = layers[params[\"type\"]](filters=int(params[\"filters\"]), kernel_size=params[\"kernel_size\"], padding=params[\"padding\"])\n",
    "                    #print(params[\"output_shape\"])\n",
    "                    lyr.out = np.zeros(params[\"output_shape\"])\n",
    "                    params[\"input_shape\"] = tuple(params[\"input_shape\"])\n",
    "                    params[\"output_shape\"] = tuple(params[\"output_shape\"])\n",
    "                if lyr_type == Dropout:\n",
    "                    lyr = layers[params[\"type\"]](prob=params[\"prob\"])\n",
    "                    try:\n",
    "                        params[\"input_shape\"] = tuple(params[\"input_shape\"])\n",
    "                        params[\"output_shape\"] = tuple(params[\"output_shape\"])\n",
    "                    except:\n",
    "                        pass\n",
    "                    \n",
    "                if lyr_type == Pool2d:\n",
    "                    lyr = layers[params[\"type\"]](kernel_size = params[\"kernel_size\"], stride=params[\"stride\"], kind=params[\"kind\"])\n",
    "                    params[\"input_shape\"] = tuple(params[\"input_shape\"])\n",
    "                    try:\n",
    "                        params[\"output_shape\"] = tuple(params[\"output_shape\"])\n",
    "                    except:\n",
    "                        pass\n",
    "                if lyr_type == Flatten:\n",
    "                    params[\"input_shape\"] = tuple(params[\"input_shape\"])                    \n",
    "                    lyr = layers[params[\"type\"]](input_shape=params[\"input_shape\"])\n",
    "                lyr.name = layer\n",
    "                lyr.activation = params[\"activation\"]\n",
    "                lyr.isbias = params[\"isbias\"]\n",
    "                lyr.input_shape = params[\"input_shape\"]\n",
    "                lyr.output_shape = params[\"output_shape\"]\n",
    "                lyr.parameters = int(params[\"parameters\"])\n",
    "                \n",
    "                if params.get(\"weights\"):\n",
    "                    lyr.weights = np.array(params[\"weights\"])\n",
    "                \n",
    "                if params.get(\"biases\"):\n",
    "                    lyr.biases = np.array(params[\"biases\"])               \n",
    "                \n",
    "                model.layers.append(lyr)\n",
    "        print(\"Model Loaded...\")\n",
    "        return model\n",
    "mm = load_model()\n",
    "mm.summary()\n",
    "m.predict(x[0]) == mm.predict(x[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-11T05:18:26.625139Z",
     "start_time": "2020-06-11T05:18:17.228475Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Input Layer\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 2160x2160 with 8 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Pool2d1\n",
      "Input\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 2160x2160 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Output\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA1UAAADhCAYAAAAgTJ84AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAdyElEQVR4nO3de5RddZnm8eehiqKSEEjJJYZECdqCAVSQjEbTOhBghNYeUEBBZYLoZNSxu2d0rQbb5ShrxqFB7bZ7tcJEhcRWLq4AA80dbZGBAWNiO3KXqyQhMZgLJmA6JLzzx9khp+61f79d51Ln+1mrVtU5dd7ze89OPXXynr3PLkeEAAAAAABp9mh2AwAAAADQzhiqAAAAACADQxUAAAAAZGCoAgAAAIAMDFUAAAAAkIGhCgAAAAAyMFS1Kdt32v5EYu2XbX9/HHp60Pax47kGMJ7IFVA9cgVUi0y1prYeqmw/bfuEZvcxEtvn2N5pe6vt39v+pe33VbzGe23fbXuz7XW2v217aon6xbYftf2y7XNGue0S29uLx7Pr40OSFBFHRMSdQ9TMth22u8s+tjJsH2V7pe0Xi89Hjed6ExW5emUNcqVyjwPDI1evrNHxubJ9qO3rbT9ne6Pt22wfNl7rTVRk6pU1yJS9v+17bG8otsO9tueP13rDaeuhqtlK/IDcGxF7S5om6buSfmj7VRW2sq+k/yHpIElzJM2S9NUS9f9P0qcl/WKMt784Ivau+7i6VLcljWU72+6RdL2k70vqk7RU0vXF9Wgj5Kp1clUo+zjQgshVS+VqmqQbJB0mabqk5ao9f6GNkKmWytRWSedKOkC1/wNeJOmfxvtFx4Em5FBlu8/2jcWrQJuKr2cV3zvD9soBt/+c7f9dfL2X7a/Zfsb2b21fantS8b1jba+2fZ7tdZIuL9NXRLws6TJJkyS9rrjP/2j78eLVqhtsH1TX1ztt/9z288Xndw5zv1dExK0R8WJEbJL0bUnz6+7nENs/tb3F9h2S9h9Q/82I+LGkbWUez0AjvGp0V/F5c/GqxjuK259r++Hi3+g22wfX3VfY/s+2H5P02BiWP1ZSt6RvRMS/RsTfS7KkBTmPCbuRq47MVWWPA0MjV52Xq4hYHhHfjYiNEfGSpL+VdJjt/XIeE2rIVEdmaltEPFpsY0vaqdpwVeXwOqoJOVSp9rgul3SwpNdK+oOkfyi+d4OkQ2zPqbv9RyX9Y/H1RZIOlXSUpD+SNFPSf6u77atV+0c6WNIiSXJtV+Mfj9aUaxPzJ1SbqB+zvUDShZI+KGmGpN9Iuqq47ask3STp7yXtJ+lvJN00xl+675b0YN3lKyStVC1I/13SwjHcR5XeXXyeVryqca/tUyX9laQPqPbKwv+RdOWAulMlvV3S4ZJU/GI8f5g1jpD0q4iIuut+VVyPapCrzssVxh+5IlfvlrQuIjZkdw+JTHVspmz/SrXh8AZJ34mI9dU9jDGIiLb9kPS0pBPGcLujJG2qu3yJpK8UXx8haZOkvVSbbl+Q9Pq6275D0lPF18dK2i6pt0SP50jaIWmzpN9Jum9Xz6rtBr647rZ7S3pJ0mxJZ0taPuC+7pV0TvH1nZI+McR6JxaP59Di8muL9afU3eYKSd8fovbuXfc/wuNZotoP7OZdj2mofw9JX961RvF4QlJ33W1vkfTxust7SHpR0sHF5ZC0oMR2/qKkqwZc9wNJX272z2m7fZArcjVMj6M+Dj7IVd33ydXYtvksSWskndXsn9F2+yBTZGqYHnslnSVpYaN/Jifknirbk23/L9u/sf171XY9TrPdVdxkqaQP27ZqP7g/jIh/VW1anixpZfHKw2ZJtxbX7/JcRJTdRXpfREyLiP0jYl5E/Ki4/iDVXpmQJEXEVkkbVHtlpN/3Cr8pvjfc456nWlhOj4hf162xKSJeGHA/Ob5WPJ5pEbH/6Dcf0sGS/q5uO29U7Rda/eNbVeL+tkraZ8B1+0jaktgfBiBXHZkrjDNy1bm5sn2ApNslfSsiBr5Kj0RkqnMzJb1yKOCVks63/ZbE/pJMyKFK0udUewPo2yNiH+3e9WhJioj7VHu14V2SPqzdu31/p9pu4iPqfmD2jdobDHepP7ws17Oq/WDVmrOnqLabd83A7xVeW3xvENtHq7a789yoHRu7y1pJfcV9199PIw21zVZJ+k9123laREyKiP87St1wHpT05uKX5C5vVv9d4MhDrnbrlFxh/JGr3TomV7b7VBuoboiIr2T0i8HI1G4dk6kh7KnivWuNMhGGqj1t99Z9dEuaqlowNhfHpX5piLrvqXaM7Y6IuFt65U2E35b0t7YPlCTbM22/Z5x6v0LSx1w7Ffhekv6npJ9FxNOSbpZ0qO0P2+527ZSVh0u6ceCd2D5StVdT/iwi/qn+exHxG0krJF1gu6c47vdPB9T32O5V7RfOru1Z5c/Gc5JeVv8f7kslfd72EUUP+9o+I2ONO1V7Y+Kfu/ZG088U1/9zxn12MnJFrlTcx3g/jk5CrsiVbO8j6TZJ90QE72fMQ6bIlGzPs/3HxWOZZPs81c6s+bOcxsuaCE+MN6sWnl0fX5b0DdXOrrLr+NVbh6j7R0lHavcrFLucJ+lxSfe5ttv4R6q94jEs185m8q6yjRevJnxR0jWqvZrweklnFt/bIOl9qr3iskHSX0p6X0T8boi7+pxqu6e/691/N6B+D82HVXuz30bVfrl8b0D97aptu3dKWlx8/W5VJCJelPQVSfe4tqt3XkRcp9obQq8qtvMDkk4e6X5s32L7r4ZZY7tqb2r8D6od53uupFOL61EeuSJXu4zr4+gw5IpcSdL7Jf0b1f5DXf/3fhq9B2EiIFNkSqq9J+6bqm2rNZL+RNJ7I+LZqh7HWDiiM48Ece0UmeslvTUixnRqYQAjI1dA9cgVUC0yhfEwEfZUpfqUpJ8TJqBS5AqoHrkCqkWmULmG/qXhVmH7adWOHT21ya0AEwa5AqpHroBqkSmMl449/A8AAAAAqtDJh/8BAAAAQDaGKgAAAADI0ND3VPX29saUKVNGvyHQwl544QVt27bNo9+yMWxzDC8mhIhoiVxNmzYtZsyY0ew2gCxr167V5s2bWyJTktTV1RXd3R35Vn5MIDt27NDOnTuHzFVDf7qnTJmik08e8TT0QMu75ZZbmt0CgHE0Y8YMXXbZZc1uA8hy7rnnNruFfrq7u3XQQQc1uw0gy7PPDv+nrzj8DwAAAAAyZA1Vtk+y/ajtx22fX1VTQCcjV0D1yBVQLTIF9Jc8VNnukvRNSSdLOlzSWbYPr6oxoBORK6B65AqoFpkCBsvZU/U2SY9HxJMRsV3SVZJOqaYtoGORK6B65AqoFpkCBsgZqmZKWlV3eXVxHYB05AqoHrkCqkWmgAFyzv431OkEB53a2fYiSYskafLkyRnLAR2hdK4AjGrUXNVnavr06Y3oCWhnpZ+rurq6xrsnoKly9lStlvSausuzJA06z2BELI6IuRExt7e3N2M5oCOUzlXDOgPa16i5qs9UX19fQ5sD2lDp5yqGKkx0OUPVzyW9wfYhtnsknSnphmraAjoWuQKqR66AapEpYIDkw/8iYoftz0i6TVKXpMsi4sHKOgM6ELkCqkeugGqRKWCwnPdUKSJulnRzRb0AELkCxgO5AqpFpoD+sv74LwAAAAB0uqw9Vdht3rx5SXVTpkxJqnv44YdL19hDnaxndKtWrRr9RkPYYw9mdjTH6aefnlQ3Z86c0jVXXHFF0lpPPPFEUh2Q4/DD0/4+6/PPP59Ut3z58tI169atS1rrqKOOSqrjBApolve///1Jde9617tK1+y1115Jaz355JNJdV//+teT6toZ/+sFAAAAgAwMVQAAAACQgaEKAAAAADIwVAEAAABABoYqAAAAAMjAUAUAAAAAGRiqAAAAACADQxUAAAAAZGCoAgAAAIAMDFUAAAAAkIGhCgAAAAAyMFQBAAAAQAaGKgAAAADI0N3sBiaKN73pTUl1U6ZMSao777zzStdMnTo1aa0999wzqW7u3LlJdUCuZcuWNbsFoCW9/PLLSXW2k+ouvPDC0jWHHnpo0lorV65Mqlu0aFFSHbDL1q1bk+o2b96cVHf//feXrjnyyCOT1jr77LOT6lasWJFU99Of/jSprhWwpwoAAAAAMjBUAQAAAECG5KHK9mts/8T2w7YftP0XVTYGdCJyBVSPXAHVIlPAYDnvqdoh6XMR8QvbUyWttH1HRDxUUW9AJyJXQPXIFVAtMgUMkLynKiLWRsQviq+3SHpY0syqGgM6EbkCqkeugGqRKWCwSt5TZXu2pKMl/ayK+wNAroDxQK6AapEpoCb7lOq295Z0jaT/EhG/H+L7iyQtkqTJkyfnLgd0hDK5AjA2I+WqPlPTp09vQndA+ynzXNXV1dXg7oDGytpTZXtP1cL0g4i4dqjbRMTiiJgbEXN7e3tzlgM6QtlcNbY7oD2Nlqv6TPX19TW+QaDNlH2uYqjCRJdz9j9L+q6khyPib6prCehc5AqoHrkCqkWmgMFy9lTNl3S2pAW2f1l8/ElFfQGdilwB1SNXQLXIFDBA8nuqIuJuSa6wF6DjkSugeuQKqBaZAgar5Ox/AAAAANCpss/+h5pnnnkmqa6npyep7rnnnmtITY65czmHAgC0kvXr1yfV3X///Ul1J554Yuman/zkJ0lrTZo0KakOyLX33nsn1V1++eVJdfPnzy9dc8899ySt9da3vjWp7oUXXkiqa2fsqQIAAACADAxVAAAAAJCBoQoAAAAAMjBUAQAAAEAGhioAAAAAyMBQBQAAAAAZGKoAAAAAIANDFQAAAABkYKgCAAAAgAwMVQAAAACQgaEKAAAAADIwVAEAAABABoYqAAAAAMjQ3ewGJorbbrstqe6EE05Iqlu4cGHpmqVLlyatBQCYGDZu3JhUd9xxxyXVbdu2rXTNTTfdlLTWzJkzk+qAZpk0aVJS3b333lu6ZsGCBUlrzZ49O6nuoIMOSqprZ+ypAgAAAIAMDFUAAAAAkIGhCgAAAAAyZA9Vtrts/4vtG6toCAC5AsYDuQKqRaaA3arYU/UXkh6u4H4A7EaugOqRK6BaZAooZA1VtmdJeq+k71TTDgByBVSPXAHVIlNAf7l7qr4h6S8lvTzcDWwvsr3C9oqUU6sCHahUrhrXFtDWRsxVfaY2bdrU2M6A9lTquWrnzp2N6wxoguShyvb7JK2PiJUj3S4iFkfE3IiY29vbm7oc0BFSctWg1oC2NZZc1Weqr6+vgd0B7Sfluaqrq6tB3QHNkbOnar6kf2/7aUlXSVpg+/uVdAV0LnIFVI9cAdUiU8AAyUNVRHw+ImZFxGxJZ0r654j4aGWdAR2IXAHVI1dAtcgUMBh/pwoAAAAAMnRXcScRcaekO6u4LwA15AqoHrkCqkWmgBr2VAEAAABAhkr2VEFatGhRUt2kSZOS6mbNmlW6ZunSpUlrAe3mbW97W1Ld8uXLK+4EaC3HHHNMUt2aNWuS6h566KHSNc8//3zSWitXjngiOmDcvPTSS0l1n/zkJ5Pqtm7dWrrmqaeeSlpr2bJlSXWdeLZH9lQBAAAAQAaGKgAAAADIwFAFAAAAABkYqgAAAAAgA0MVAAAAAGRgqAIAAACADAxVAAAAAJCBoQoAAAAAMjBUAQAAAEAGhioAAAAAyMBQBQAAAAAZGKoAAAAAIANDFQAAAABk6G52AxPFHnukzafLli1Lqrv44ouT6oBOsHz58ma3ALSkX//610l1a9asSap76KGHSteceuqpSWs99thjSXVArh07diTVXXrppUl18+fPT6pLcdpppyXVbdy4seJOWh97qgAAAAAgA0MVAAAAAGTIGqpsT7O9zPYjth+2/Y6qGgM6FbkCqkeugGqRKaC/3PdU/Z2kWyPidNs9kiZX0BPQ6cgVUD1yBVSLTAF1kocq2/tIerekcyQpIrZL2l5NW0BnIldA9cgVUC0yBQyWc/jf6yQ9J+ly2/9i+zu2p1TUF9CpyBVQPXIFVItMAQPkDFXdkt4q6ZKIOFrSC5LOH3gj24tsr7C9Ytu2bRnLAR2hdK4a3SDQhkbNVX2mNm3a1IwegXZS+rlq586dje4RaKicoWq1pNUR8bPi8jLVAtZPRCyOiLkRMbe3tzdjOaAjlM5VQ7sD2tOouarPVF9fX8MbBNpM6eeqrq6uhjYINFryUBUR6yStsn1YcdXxksr/lT8AryBXQPXIFVAtMgUMlnv2vz+T9IPirC9PSvpYfktAxyNXQPXIFVAtMgXUyRqqIuKXkjj8CKgQuQKqR66AapEpoL+sP/4LAAAAAJ0u9/A/FK699tqGrjd79uzSNQsXLkxa64ILLkiq27497U9W9PT0JNUBAEaW+vt15syZSXUf+1j5I8Le/va3J6316U9/Oqluy5YtSXVTp05NqsPEM2nSpKQ620l1P/rRj5LqUtx1111JdW984xsr7qT1sacKAAAAADIwVAEAAABABoYqAAAAAMjAUAUAAAAAGRiqAAAAACADQxUAAAAAZGCoAgAAAIAMDFUAAAAAkIGhCgAAAAAyMFQBAAAAQAaGKgAAAADIwFAFAAAAABkYqgAAAAAgQ3ezGxhPM2bMKF3T3Z22SVatWpVUl+qiiy4qXfPBD34waa0LLrggqW7btm1JdT09PUl1aF2f+tSnkuouueSSijsBWs8RRxxRuubBBx9MWmvDhg1JdanmzZtXuubAAw9MWuu6665Lqjv//POT6nbu3JlUh9Z1/PHHJ9U9+eSTSXVPPfVUUt3s2bNL16xZsyZpraeffjqpbsuWLUl17Yw9VQAAAACQgaEKAAAAADIwVAEAAABAhqyhyvZ/tf2g7QdsX2m7t6rGgE5FroDqkSugWmQK6C95qLI9U9KfS5obEUdK6pJ0ZlWNAZ2IXAHVI1dAtcgUMFju4X/dkibZ7pY0WdKz+S0BHY9cAdUjV0C1yBRQJ3moiog1kr4m6RlJayU9HxG3D7yd7UW2V9hekXqKbaBTpOSq0T0C7WYsuarP1KZNm5rRJtA2Up6rOAU9Jrqcw//6JJ0i6RBJB0maYvujA28XEYsjYm5EzO3t5XBbYCQpuWp0j0C7GUuu6jPV19fXjDaBtpHyXNXV1dXoNoGGyjn87wRJT0XEcxHxkqRrJb2zmraAjkWugOqRK6BaZAoYIGeoekbSPNuTbVvS8ZIerqYtoGORK6B65AqoFpkCBsh5T9XPJC2T9AtJ9xf3tbiivoCORK6A6pEroFpkChisO6c4Ir4k6UsV9QJA5AoYD+QKqBaZAvrLPaU6AAAAAHS0rD1Vre4jH/lI6Zrvfe97SWtt3bo1qe7oo49Oqps3b17pmq9+9atJa6XaZ599GroeWte3vvWtpLpLLrkkqe5DH/pQUt3VV1+dVAfkSPld+dJLLyWttW7duqS6t7zlLUl1Kaenv/zyy5PWSs3vyy+/nFSH1rZ27drSNR/4wAeS1vrDH/6QVHfVVVcl1R1wwAGla3p6epLW2rBhQ1LdjTfemFTXzthTBQAAAAAZGKoAAAAAIANDFQAAAABkYKgCAAAAgAwMVQAAAACQgaEKAAAAADIwVAEAAABABoYqAAAAAMjAUAUAAAAAGRiqAAAAACADQxUAAAAAZGCoAgAAAIAMDFUAAAAAkKG72Q2Mp7vvvrt0zc0335y01pw5c5Lq9t1336S6Sy+9tHTNhRdemLTWGWeckVQH7GK7oetdffXVDV0PyPHQQw+VrnnmmWeS1tq5c2dSXWqmnnjiidI19913X9Ja11xzTVLd1q1bk+rQ2vbbb7/SNan/T/riF7+YVHfaaacl1aX8zti+fXvSWrfffntSXervqHbGnioAAAAAyMBQBQAAAAAZRh2qbF9me73tB+que5XtO2w/VnzuG982gYmFXAHVI1dA9cgVMDZj2VO1RNJJA647X9KPI+INkn5cXAYwdktEroCqLRG5Aqq2ROQKGNWoQ1VE3CVp44CrT5G0tPh6qaRTK+4LmNDIFVA9cgVUj1wBY5P6nqrpEbFWkorPB1bXEtCxyBVQPXIFVI9cAQOM+ynVbS+StEiSJk+ePN7LAR2hPlcA8tVnavr06U3uBpgY6nPV1dXV5G6A8ZW6p+q3tmdIUvF5/XA3jIjFETE3Iub29vYmLgd0hKRcNaw7oD2NKVf1merr4z33wChK54qhChNd6lB1g6SFxdcLJV1fTTtARyNXQPXIFVA9cgUMMJZTql8p6V5Jh9lebfvjkv5a0om2H5N0YnEZwBiRK6B65AqoHrkCxmbU91RFxFnDfOv4insBOga5AqpHroDqkStgbFIP/wMAAAAAqAFn/2umNWvWlK45+eSTk9Zavnx5Ut0XvvCFpLoXX3yxdM1xxx2XtFZPT09SHQBgdJMmTSpdc8wxxySt9cgjjyTVve51r0uq23PPPUvXfPazn01aa8uWLUl1mJhS/u+yevXqpLXe8573JNWlmjFjRumaBQsWJK316KOPJtV1IvZUAQAAAEAGhioAAAAAyMBQBQAAAAAZGKoAAAAAIANDFQAAAABkYKgCAAAAgAwMVQAAAACQgaEKAAAAADIwVAEAAABABoYqAAAAAMjAUAUAAAAAGRiqAAAAACADQxUAAAAAZOhudgPjadWqVQ1b65BDDmloHQBgYli3bl3D1nr1q1/dsLUk6fWvf33pmhdffHEcOgHGz+zZs5vdwqjuueeeZrcw4bGnCgAAAAAyMFQBAAAAQIZRhyrbl9leb/uBuuu+avsR27+yfZ3taePbJjCxkCugeuQKqBaZAsZuLHuqlkg6acB1d0g6MiLeLOnXkj5fcV/ARLdE5Aqo2hKRK6BKS0SmgDEZdaiKiLskbRxw3e0RsaO4eJ+kWePQGzBhkSugeuQKqBaZAsauivdUnSvplgruB8Bu5AqoHrkCqkWmgELWKdVtf0HSDkk/GOE2iyQtkqTJkyfnLAd0hLK5AjC60XJVn6np06c3sDOgPZV9rurq6mpQZ0BzJA9VthdKep+k4yMihrtdRCyWtFiS9ttvv2FvByAtV7bJFTCCseSqPlNz5swhU8AIUp6r9tprL3KFCS1pqLJ9kqTzJP3biOCv9AEVIFdA9cgVUC0yBQxtLKdUv1LSvZIOs73a9scl/YOkqZLusP1L25eOc5/AhEKugOqRK6BaZAoYu1H3VEXEWUNc/d1x6AXoGOQKqB65AqpFpoCxq+LsfwAAAADQsRiqAAAAACCDRzhpS/WL2c9J+s0w395f0u8a1kzrY3sM1irb5OCIOKDZTexCrkphe/TXStujZXJFpkpjm/TXKtujZTIlkauS2B6Dtco2GTZXDR2qRmJ7RUTMbXYfrYLtMRjbpDy2WX9sj/7YHuWxzQZjm/TH9iiPbdYf22OwdtgmHP4HAAAAABkYqgAAAAAgQysNVYub3UCLYXsMxjYpj23WH9ujP7ZHeWyzwdgm/bE9ymOb9cf2GKzlt0nLvKcKAAAAANpRK+2pAgAAAIC20/ShyvZJth+1/bjt85vdTyuw/bTt+23/0vaKZvfTDLYvs73e9gN1173K9h22Hys+9zWzx1ZGrgbr9FyRqXzkqr9Oz5RErnKRqcHIVfvmqqlDle0uSd+UdLKkwyWdZfvwZvbUQo6LiKNa/fSR42iJpJMGXHe+pB9HxBsk/bi4jAHI1Yg6OVdLRKaSkathdXKmJHKVjEyNiFy1Ya6avafqbZIej4gnI2K7pKskndLkntACIuIuSRsHXH2KpKXF10slndrQptoHucIgZCobucIg5CoLmcKQ2jVXzR6qZkpaVXd5dXFdpwtJt9teaXtRs5tpIdMjYq0kFZ8PbHI/rYpcDY1cDUamxo5cDUamhkauxoZMDY1cDa3lc9Xd5PU9xHWcjlCaHxHP2j5Q0h22HymmdmAsyNXQyBVykKvByBRykKmhkas21ew9Vaslvabu8ixJzzapl5YREc8Wn9dLuk61XeSQfmt7hiQVn9c3uZ9WRa6GQK6GRKbGjlwNQKaGRa7GhkwNgVwNq+Vz1eyh6ueS3mD7ENs9ks6UdEOTe2oq21NsT931taR/J+mBkas6xg2SFhZfL5R0fRN7aWXkagByNSwyNXbkqg6ZGhG5GhsyNQC5GlHL56qph/9FxA7bn5F0m6QuSZdFxIPN7KkFTJd0nW2p9u9zRUTc2tyWGs/2lZKOlbS/7dWSviTpryX90PbHJT0j6Yzmddi6yNWQOj5XZCoPuRqk4zMlkascZGpI5ErtmytHcPgqAAAAAKRq9uF/AAAAANDWGKoAAAAAIANDFQAAAABkYKgCAAAAgAwMVQAAAACQgaEKAAAAADIwVAEAAABABoYqAAAAAMjw/wEXxQNXc1vVSAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 2160x2160 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Conv2d2\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 2160x2160 with 16 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dropout3\n",
      "Input\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 2160x2160 with 8 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Output\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 2160x2160 with 8 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Flatten4\n",
      "Out Layer(FFL)\n"
     ]
    }
   ],
   "source": [
    "timg = x[10]\n",
    "plt.imshow(timg.reshape(28, 28))\n",
    "op = mm.predict(timg)\n",
    "for l in mm.layers:\n",
    "    print(l.name)\n",
    "    if type(l).__name__ == \"Conv2d\":\n",
    "        fig = plt.figure(figsize=(30, 30))\n",
    "        cols = l.filters * 2\n",
    "        rows = 1\n",
    "        f = 0\n",
    "        for i in range(0, cols*rows):\n",
    "            fig.add_subplot(rows, cols, i+1)                    \n",
    "            if i % 2 == 0:\n",
    "                if f < l.filters:\n",
    "                    plt.imshow(l.out[:, :, f], cmap=\"gray\")\n",
    "                   \n",
    "            else:\n",
    "                if f < l.filters:\n",
    "                    cimg = l.weights[:, :, 0, f]\n",
    "                    plt.imshow(cimg, cmap='gray')\n",
    "                    plt.title(f\"Layer: {l.name} Filter: {f}\")\n",
    "                    f+=1                    \n",
    "        #plt.show()\n",
    "    if type(l).__name__ == \"Pool2d\":\n",
    "        fig = plt.figure(figsize=(30, 30))\n",
    "        cols = l.out.shape[2] * 2\n",
    "        rows = 1\n",
    "        print(\"Input\\n\")\n",
    "        for f in range(l.out.shape[2]):\n",
    "            fig.add_subplot(rows, cols, f+1)                    \n",
    "            plt.imshow(l.input[:, :, f], cmap=\"gray\")\n",
    "            plt.title(f\"Layer: {l.name} Filter: {f}\")\n",
    "        plt.show()\n",
    "        fig = plt.figure(figsize=(30, 30))\n",
    "        print(\"Output\\n\")\n",
    "        for f in range(l.out.shape[2]):\n",
    "            fig.add_subplot(rows, cols, f+1)                    \n",
    "            plt.imshow(l.out[:, :, f], cmap=\"gray\")\n",
    "            plt.title(f\"Layer: {l.name} Filter: {f}\")\n",
    "    if type(l).__name__ == \"Dropout\":\n",
    "        try:\n",
    "            fig = plt.figure(figsize=(30, 30))\n",
    "            cols = l.out.shape[2] * 2\n",
    "            rows = 1\n",
    "            print(\"Input\\n\")\n",
    "            for f in range(l.out.shape[2]):\n",
    "                fig.add_subplot(rows, cols, f+1)                    \n",
    "                plt.imshow(l.input[:, :, f], cmap=\"gray\")\n",
    "                plt.title(f\"Layer: {l.name} Filter: {f}\")\n",
    "            plt.show()\n",
    "            fig = plt.figure(figsize=(30, 30))\n",
    "            print(\"Output\\n\")\n",
    "            for f in range(l.out.shape[2]):\n",
    "                fig.add_subplot(rows, cols, f+1)                    \n",
    "                plt.imshow(l.out[:, :, f], cmap=\"gray\")\n",
    "                plt.title(f\"Layer: {l.name} Filter: {f}\")\n",
    "        except:\n",
    "            pass\n",
    "    plt.show()\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-11T04:04:14.344052Z",
     "start_time": "2020-06-11T04:04:10.373340Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Input Layer\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 2160x720 with 8 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Conv2d2\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 2160x720 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 2160x720 with 16 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 2160x720 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 2160x720 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 2160x720 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(timg.reshape(28, 28), cmap=\"gray\")\n",
    "plt.show()\n",
    "\n",
    "for l in mm.layers:\n",
    "    fig = plt.figure(figsize=(30, 10))\n",
    "    if type(l).__name__ == \"Conv2d\":\n",
    "        print(l.name)\n",
    "        cols = l.filters * 2\n",
    "        rows = 1\n",
    "        f = 0\n",
    "        for i in range(0, cols*rows):\n",
    "            fig.add_subplot(rows, cols, i+1)                    \n",
    "            if i %2 == 0:\n",
    "                if f < l.filters:\n",
    "                    plt.imshow(l.weights[:, :, c, f], cmap=\"gray\")\n",
    "                    plt.title(f\"Layer: {l.name} Filter: {f} Channel: {c}\")\n",
    "            else:\n",
    "                if f < l.filters:\n",
    "                    cimg = cv2.filter2D(timg, -1, l.weights[:, :, 0, f])\n",
    "                    plt.imshow(cimg, cmap='gray')\n",
    "                    f+=1\n",
    "                    \n",
    "        plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-11T03:26:14.439935Z",
     "start_time": "2020-06-11T03:26:05.086857Z"
    },
    "code_folding": [],
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAQ8AAAEICAYAAABBKnGGAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAVaklEQVR4nO3de7QdZX3G8e/TXLknEJEQCIESL0BVYhYXL6ysghVSJbSiRRQIBSMoFRWUiC0i1haplZYFmgZBCSCgQCHYCAWBImoogSaEkCIBwcSEewgEInjw1z/mPWXY2eecfd49+3Lg+ay115nLO/P+9szOs2dmz4AiAjOzwfqjThdgZkOTw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8rC0kTZS0XtKwNH6rpGM7XVcuSadLuqTTddRq53btmvCQ9LCkAzpdR38kzZR0e5v6Ckm7dkMtg5H244YUFL2v7SPiNxGxeUS8XGeZlr8XSX8u6XZJz0h6VNL5krYYYJnDJS1K72GNpJ9Iek8r62w3SZ9L22OdpAsljWp02a4Jj06TNLzTNQwlKvT1+flgCore1+oW19LIvtsK+Htge+CtwA7AP/Wzzs8D/wL8A/BGYCLwbWBGs/V2C0nvB2YD+wOTgF2Arza6fNeHh6Sxkn4s6QlJa9PwDmnehyXdVdP+JEnXpOFRkr4p6TeSHpM0R9Imad40SasknSLpUeB7GbU9LOlkSfek5L5C0uia9Z8q6cnU9mOlZV91eFn+9pV0W5q8JH3r/dUg6zpa0nJJz0l6SNInS/PulfTB0viIVN870vg+kn6RvqGXSJpWU/PXJf0ceIHiw9ZoTZPS0dTwmulvBeYA+6b3+kyaXum+i4gfRMT1EfFCRKwFzgfe3UetWwFnAJ+OiKsj4vmI+H1EXBcRXyg1HSlpXtrOyyRNLa1jtqQH07z7JP1Fad7MdBT0zfSZ/rWkg0rzb5X0NUk/T8v/p6Rxpfl97qNBOgq4ICKWpW3yNWBmowt3fXhQ1Pg9YCeK9N8AnJvmzQd2Th/AXh8HLk7D3wDeBLwD2BWYAJxWarsdsHVa9yyAtEMGc2j6EeBAYGfgbbx6428HjEv9HgXMlfTmgVYYEfulwbenb+4rBlEPwOPAB4AtgaOBsyVNSfPmUWyjXtOBNRGxWNIE4D8ovqG3Bk4GrpL0hlL7Iyi21RbAI4OsayMRsRw4Dvhleq9j0qxW77v9gGV9zNsXGA38+wDrOBi4HBhD8Vk8tzTvQeC9FEc8XwUukTS+NH9v4H6Kz8dZwAWSVJp/OMW+2xYYSbEvaHAfkdpOTNtkYh/17w4sKY0vAd4oaZsB3nchIrriBTwMHNBAu3cAa0vj3wG+noZ3B9YCowABzwN/XGq7L/DrNDwNeAkYPYgaZwK319T88dL4WcCc0vp7gM1K838I/F0avhU4tp91B7Bro7UMUPc1wIlpeHvgOWDLNH4l8MU0fApwcc2yNwBHlWo+o4H9uB54Jr2uSdMnpfc0vPb913nvle+7mhrflz4nb+pj/seARwdYx+nATaXx3YAN/bRfDMwovd8VpXmbpm2zXWnb/G1p/qeA6wexj47tr/bScg8CB5bGR6Q6JjWyfNcfeUjaVNK/SXpE0rPAbcAYpav2wEXA4Sm1jwB+GBEvAm+g2Cl3pfR9Brg+Te/1RET8rskSHy0NvwBsXhpfGxHPl8YfofjH21KSDpK0UNLT6X1Pp/iGI4rrDz8HPiRpDHAQcGladCfgw73bKy37HqD8jbmygRIOiYgx6XVIxlto2b6TtA/wA+DQiPhVH82eAsbVnmLVUbvvR/cuI+lISYtL9e9B2ge1y0bEC2lw83rzefXnqpF91Kj1FEenvXqHn2tk4aFwkfAk4M3A3hHxaDo3/x+KbyciYqGklygOEQ9PL4AnKU5xdo+I3/ax7lY/UjxW0malAJkI3JuGn6f4B9Jruyo6VHG1/CrgSODaiPi9imtA5UPii4BjKfb/L0vbZyXFt9on+umiFdusdp0t2XeS9qQ4vfjriPhpP01/CfwOOITiyGyw/exEcU1lf4rt+7Kkxbx6H+RqZB81ahnwdoojYtLwYxHxVCMLd9uRxwhJo0uv4RTn1huAZyRtDXylznLzKM43eyLidoCI+APFDjxb0rZQnC+quMLcTl+VNFLSeymuQ/woTV8M/GU6stoVOKZmuccY+IKkarbXaIrz41HAE0BPuhD3ZzXLXQNMAU6k2Ha9LgE+KOn9koaldU5TukDdQo8BO0gaCa3Zd5L2oDh6+ZuIuK6/thGxjuL6ynmSDkn7aEQ6ojurge42owi3J1LfR1MceVShyn00DzhG0m6SxgJ/C3y/0YW7LTwWUARF7+t0ip/LNqH4NlpI8QGodTHFzrm4ZvopwApgYTrluYniKKZPKq74vzf/LbzKoxTn1qspTg2Oi4j/TfPOpjhvf4ziSODSmmVPBy5Kh6Yf6WP97+LV26v39RmKb5O1FEdi88sLRcQGiqOTnYGrS9NXUvwUeSrFB38l8AVa/zm5meJb8FFJT6ZpVe+7kyhOey7QK/ef9HXBlIj4FvB5in9QvdviBIrg7VdE3Af8M8URzGPAn1CcKjZtMPtIr9yYV/eCaURcT3Gd7haKU+pHqP/lXJfShZIhTcVPeI8DUyLigU7XA8XPicAlEdHqb+0skk6juGD48QEbm9UxFK55NOJ44M5uCY5ul07/jqG4wGyWpanwSB/CKyh+hnsY+EgUN5vUtnsZWJpGfxMRBzfTb826H6a4EJVzVf91R9InKE4FL46I2wZqb9aXpk5b0sWjpyPiTEmzgbERcUqddusjYvON12BmQ1Wz4XE/MC0i1qS7526NiI0uajk8zF57mg2PZ+KV24mRtDYixtZp10Px02QPcGZE1L1iLWkW6VbjYaOHv3PznTZalSV/uL+n0yV0vS133+gBXqvx22XPPhkRG93a3ogBr3lIuon6NzB9eRD9TIyI1ZJ2AW6WtDQiHqxtFBFzgbkAY96ybUz77qGD6OL15fn9nuh0CV3vgB82dKPk69rs3W/Ifj5pwPCIiD7/GxsqnnYcXzptebyPdaxOfx+SdCuwJ8V99WY2RDV78898iqdFSX+vrW2g4pH6UWl4HMVj0Pc12a+ZdViz4XEm8D5JD1A8qXgmgKSpkr6b2rwVWCRpCcWdbGemO/DMbAhr6j6P9ADN/nWmL6J48IqI+AXF7blm9hrSbc+2mNkQ4fAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPLUkl4SDpQ0v2SVkiaXWf+KElXpPl3SJpURb9m1jlNh4ekYcB5wEHAbsBHJe1W0+wYYG1E7AqcDXyj2X7NrLOqOPLYC1gREQ9FxEvA5cCMmjYzgIvS8JXA/pJUQd9m1iFVhMcEYGVpfFWaVrdNRPQA64BtKujbzDqkivCodwQRGW2QNEvSIkmLXnpmQwWlmVmrVBEeq4AdS+M7AKv7aiNpOLAV8HTtiiJibkRMjYipI8dsUkFpZtYqVYTHncBkSTtLGgkcBsyvaTMfOCoNHwrcHBEbHXmY2dAxvNkVRESPpBOAG4BhwIURsUzSGcCiiJgPXABcLGkFxRHHYc32a2ad1XR4AETEAmBBzbTTSsO/Az5cRV9m1h18h6mZZXF4mFkWh4eZZXF4mFkWh4eZZXF4mFkWh4eZZXF4mFkWh4eZZXF4mFkWh4eZZXF4mFkWh4eZZXF4mFkWh4eZZXF4mFkWh4eZZXF4mFkWh4eZZXF4mFkWh4eZZXF4mFkWh4eZZXF4mFkWh4eZZXF4mFkWh4eZZXF4mFmWSsJD0oGS7pe0QtLsOvNnSnpC0uL0OraKfs2sc4Y3uwJJw4DzgPcBq4A7Jc2PiPtqml4RESc025+ZdYcqjjz2AlZExEMR8RJwOTCjgvWaWRdr+sgDmACsLI2vAvau0+5DkvYDfgV8LiJW1jaQNAuYBTBii7E89KPJFZT32rR49Y2dLqHrTfna8Z0uYQi4IXvJKo48VGda1IxfB0yKiLcBNwEX1VtRRMyNiKkRMXX4JptVUJqZtUoV4bEK2LE0vgOwutwgIp6KiBfT6PnAOyvo18w6qIrwuBOYLGlnSSOBw4D55QaSxpdGDwaWV9CvmXVQ09c8IqJH0gkUJ0/DgAsjYpmkM4BFETEf+Iykg4Ee4GlgZrP9mllnVXHBlIhYACyomXZaafhLwJeq6MvMuoPvMDWzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLJWEh6QLJT0u6d4+5kvSOZJWSLpH0pQq+jWzzqnqyOP7wIH9zD8ImJxes4DvVNSvmXVIJeEREbcBT/fTZAYwLwoLgTGSxlfRt5l1RruueUwAVpbGV6VpryJplqRFkhb1bHi+TaWZWY52hYfqTIuNJkTMjYipETF1+CabtaEsM8vVrvBYBexYGt8BWN2mvs2sBdoVHvOBI9OvLvsA6yJiTZv6NrMWGF7FSiRdBkwDxklaBXwFGAEQEXOABcB0YAXwAnB0Ff2aWedUEh4R8dEB5gfw6Sr6MrPu4DtMzSyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCxLJeEh6UJJj0u6t4/50yStk7Q4vU6rol8z65xK/kfXwPeBc4F5/bT5WUR8oKL+zKzDKjnyiIjbgKerWJeZDQ1VHXk0Yl9JS4DVwMkRsay2gaRZwCyA0SO2YvzPnmljeUPLLpM/2ekSut7lXzyn0yV0vX2/nb9su8LjbmCniFgvaTpwDTC5tlFEzAXmAmy16fbRptrMLENbfm2JiGcjYn0aXgCMkDSuHX2bWWu0JTwkbSdJaXiv1O9T7ejbzFqjktMWSZcB04BxklYBXwFGAETEHOBQ4HhJPcAG4LCI8GmJ2RBWSXhExEcHmH8uxU+5ZvYa4TtMzSyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCxL0+EhaUdJt0haLmmZpBPrtJGkcyStkHSPpCnN9mtmnVXF/+i6BzgpIu6WtAVwl6QbI+K+UpuDgMnptTfwnfTXzIaopo88ImJNRNydhp8DlgMTaprNAOZFYSEwRtL4Zvs2s86p9JqHpEnAnsAdNbMmACtL46vYOGDMbAipLDwkbQ5cBXw2Ip6tnV1nkaizjlmSFkla9FLPC1WVZmYtUEl4SBpBERyXRsTVdZqsAnYsje8ArK5tFBFzI2JqREwdOXzTKkozsxap4tcWARcAyyPiW300mw8cmX512QdYFxFrmu3bzDqnil9b3g0cASyVtDhNOxWYCBARc4AFwHRgBfACcHQF/ZpZBzUdHhFxO/WvaZTbBPDpZvsys+7hO0zNLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyNB0eknaUdIuk5ZKWSTqxTptpktZJWpxepzXbr5l11vAK1tEDnBQRd0vaArhL0o0RcV9Nu59FxAcq6M/MukDTRx4RsSYi7k7DzwHLgQnNrtfMupsiorqVSZOA24A9IuLZ0vRpwFXAKmA1cHJELKuz/CxgVhrdA7i3suKqMQ54stNFlLie/nVbPdB9Nb05IrbIWbCy8JC0OfBfwNcj4uqaeVsCf4iI9ZKmA/8aEZMHWN+iiJhaSXEV6baaXE//uq0e6L6amqmnkl9bJI2gOLK4tDY4ACLi2YhYn4YXACMkjauibzPrjCp+bRFwAbA8Ir7VR5vtUjsk7ZX6farZvs2sc6r4teXdwBHAUkmL07RTgYkAETEHOBQ4XlIPsAE4LAY+X5pbQW1V67aaXE//uq0e6L6asuup9IKpmb1++A5TM8vi8DCzLF0THpK2lnSjpAfS37F9tHu5dJv7/BbUcaCk+yWtkDS7zvxRkq5I8+9I97a0VAM1zZT0RGm7HNvCWi6U9LikuvfgqHBOqvUeSVNaVcsgamrb4xENPq7R1m3UskdIIqIrXsBZwOw0PBv4Rh/t1rewhmHAg8AuwEhgCbBbTZtPAXPS8GHAFS3eLo3UNBM4t037aT9gCnBvH/OnAz8BBOwD3NEFNU0Dftym7TMemJKGtwB+VWd/tXUbNVjToLdR1xx5ADOAi9LwRcAhHahhL2BFRDwUES8Bl6e6ysp1Xgns3/szdAdrapuIuA14up8mM4B5UVgIjJE0vsM1tU009rhGW7dRgzUNWjeFxxsjYg0UbxbYto92oyUtkrRQUtUBMwFYWRpfxcYb+f/bREQPsA7YpuI6BlsTwIfSIfCVknZsYT0DabTedttX0hJJP5G0ezs6TKe0ewJ31Mzq2DbqpyYY5Daq4j6Phkm6CdiuzqwvD2I1EyNitaRdgJslLY2IB6upkHpHELW/ZTfSpkqN9HcdcFlEvCjpOIojoz9tYU39aff2acTdwE7xyuMR1wD9Ph7RrPS4xlXAZ6P0nFfv7DqLtHwbDVDToLdRW488IuKAiNijzuta4LHeQ7f09/E+1rE6/X0IuJUiRauyCih/a+9A8SBf3TaShgNb0dpD5gFrioinIuLFNHo+8M4W1jOQRrZhW0WbH48Y6HENOrCNWvEISTedtswHjkrDRwHX1jaQNFbSqDQ8juLu1tr/bkgz7gQmS9pZ0kiKC6K1v+iU6zwUuDnSFacWGbCmmvPlgynOaTtlPnBk+kVhH2Bd7+lop7Tz8YjUT7+Pa9DmbdRITVnbqB1XoBu8IrwN8FPggfR36zR9KvDdNPwuYCnFLw5LgWNaUMd0iqvRDwJfTtPOAA5Ow6OBHwErgP8GdmnDthmopn8ElqXtcgvwlhbWchmwBvg9xTfoMcBxwHFpvoDzUq1Lgalt2D4D1XRCafssBN7VwlreQ3EKcg+wOL2md3IbNVjToLeRb083syzddNpiZkOIw8PMsjg8zCyLw8PMsjg8zCyLw8PMsjg8zCzL/wEl2HVo2GzEVQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAQ8AAAEICAYAAABBKnGGAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAVc0lEQVR4nO3debBcZZ3G8e9DEsKSAFlQYhJAHFSW0REyLEGdzLAIFBJHUIMOiyWVAaUUC2sMYIEyIshYKAwOMQhFQAbC4mDAuICAqEOAwLCFiAQGzZ1EIAkkRNbAb/4479VDp/vevm+f290Xnk9V1z3d5+3z/u7b5zz3bJ0oIjAzG6iNOl2AmQ1NDg8zy+LwMLMsDg8zy+LwMLMsDg8zy+LweJORtL2kkDS84uWeIun7g9lHO0l6QtJ+na6jrNvGdUiERzd+kPVImiDpYkkrJD0n6beSviZp80Hs8y2SrpS0XNIaSb+RtOcA3v8tSY+W6j2qj7bTJL0maV3pcQNARHwjIo5t8L7bJNWdVxVJi2vqWt9bW4P2W0j6jqQ/pPZL0/Pxg1lnO0n6QVoX10r6XdWfwZAIj05rJukljQXuADYF9o6I0cD+wFbAOwaxvFHA3cDuwFhgLvBjSaOafP+fgA8DWwJHA+dJmtpH++URMar0+HALtTdF0rD+2kTELr01AaOBPwDXNFjexsAvgF2AA4EtgKnAKmCPquruAmcB20fEFsChwNcl7V7Z0iOi6x/AE8B+dV4fA9wIPA08k6YnpXkfA+6paX8ScH2aHgl8i2IlexKYDWya5k0DeoAvA38ELm+ixq8DDwIb9dFmKsWGvib9nFqadxvwr8BvgOeAnwPj07yfAifULOt+4KMN+lkL7J6mh6XfcyXwOPA5IIDhDd47HzipwbxpQE+DeV8FfpCmt+/tAzgTeBV4EVgHXJDavBu4CVgNPAJ8vLSsS4ELgQUU4bbBZ9/PZ/F3qa/NG8w/Nn3mo/pZ574EPJA+r3nAJv2td018lr1jc3Ra91YCp5beuxEwC3iMIsyuBsbWjmvGNvQuYEV5nFveLqta0GA+aBwe44DDgM0o/tpcw+vDYTWwU6n9/wCHpenvpA1lbHrvDcBZpY1kPfDNtJzeUHkWeH+DGhcCX+vjdxibVrQj00Z1RHo+rrTCPQa8k2Lv5Tbg7DTvKOA3pWXtnGoZWaefv6HYULdMz48DfgtMTjXc2mgFTP2uAA5s8DtMY4DhUfrdji213RxYBnw6jcVuaSPaJc2/lGKD3SdtTJsAnwQeaHJ9uQS4tI/5VwFzm1jn7gLelsZtCXBcf+tdE59l79hclOa9F3iJtJ4CJ6Z1aVJa974HXNlgXGcBN/bze/wH8Hx63730EZgD3i7bsfG3XGSD8Giw4TxTen4hcGaa3oViYx0JiOIv2jtKbfcG/re0kbxM+kvTZI2P9q5cDeYfCdxV89odwDGlFe4rpXmfBX6apkenerdLz88ELqnTxxYUez8nl167pVwXcACNw2MuxV6OGvwO04DXKIKr9/HxNO+rNB8enwB+VbPs7wGnp+lLgcsy15XNKPa8pvXR5ibSxtzPOvdPpefnALObXO/6+ix7x6a8p3IXMCNNLwH2Lc2bALxCEbKvG9cBjMkw4P3AV4AROeNa7zGkz3lI2kzS9yT9XtJa4HZgq9Ix8lzgk5JEsfFeHREvAVtTrGT3SHpW0rMUG83WpcU/HREvDqCcVRQfdCNvA35f89rvgYml538sTT9PcT6DiHgO+DEwI82bAVxRXpCkTSn2nhZGxFk1/S6r6XMDkv4N2JUiDPr6tuTyiNiq9Li6j7aNbAfs2Tv2afw/BWxTarOs/lv79VGKPc5f9tGmv8+qV93Po4n1ruF7m5i/HfBfpXFZQnHY99Ym6q0rIl6NiF9T7M0cn7ucWkM6PCjOYbwL2DOKk0IfTK8LICIWUuxBfIBit/fyNH8l8ALFbnLvRrBlFCfbeg3068Y3A/8oqdGYLqdYMcq2Bf6vyeVfCRwhaW+K3d1be2dIGglcn5b1zzXvW0FxyFLu83UkfQ04CDggItY2Wc9A1I7lMuCXNSE0KiKO7+M9zTqaYq+lr/ffDHyohatgfa53LVoGHFQzNptERLPrSV+GU+HJ+6EUHiMkbVJ6DKfYnX8BeDZd7Ti9zvsuAy4A1qf0JSJeozjm/LaktwBImijpQy3Udy7FYcNcSduVlnmupPdQnPx7p6RPShou6RMU5y5ubHL5CyjC5wxgXvodkDQCuJZiHI7qfb3kauDzkiZJGkNxnPxnkk6mCNb9I2LVwH/tpjwJ7FB6fiPFWBwpaUR6/K2knVrpRNIk4O8p9jj7cjnFRnqdpHdL2kjSuHSvysFNdNXMepdrNnBmaR3aWtL0gS4kXcKfIWmUpGFp3T6C4jC2EkMpPBZQfGC9j69SnPTclGJPYiHFoUetyyl2xy+vef3LwFJgYdr1vJnir0lD6X6AD9SbFxGrKa6mvALcKek5isuBa4ClacM8hOKv1irgX4BDImJln7/1X5b/EvBDYD/gP0uzpqblHkCxMq+rqfMi4GcUV2fuTcso+wbF3sijpfee0kxNA3AecLikZySdnw7DDqA4/FpOsQvfe3K6LkmfkrS4n36OBO6IiMf6apTGcj+KE8k3UZwjuQsYD9zZxO/TzHqX6zyKE/k/T+vQQqDufTsp7H7SYDlBcYjSQ3Gu71vAiRHxo6oKVd97d0NfOhfwFLBbRDza6XrM3iiG0p5HruOBux0cZtVq6R75dLw3j+IS0hMUZ+qfqdPuVYpLiAB/iIhDW+l3APU9QXES6yPt6M/szaSlwxZJ5wCrI+JsSbOAMRHx5Trt1tVcyTCzIa7V8HiE4macFZImALdFxAYnHR0eZm88rYbHsxGxVen5MxExpk679cB9FLd8nx0R1zdY3kxgJsAwhu2+GVtk1/ZGt2lLFzXfHF6Jfr9P96b3zG9XroyIrftvuaFmvi16M6+/86/XqQPoZ9uIWC5pB+AWSQ/Wu5wWEXOAOQBbaGzsqX0H0MWby05XdMU/6dDVnnpxdKdL6Hrzpl5U947jZvS7BkZEw39HQ9KTkiaUDluearCM5enn45JuA95H8cUhMxuiWr1UO5/idmDSzw1uQJE0Jt0+TfqHVvYBHm6xXzPrsFbD42xgf0mPUvzDN2cDSJqi9E/SATsBiyTdT/F9jLMjwuFhNsS1dOCcbrne4MRERCyi+AdXiIj/Bv66lX7MrPu8Ge4wNbNB4PAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPLUkl4SDpQ0iOSlkqaVWf+SEnz0vw7JW1fRb9m1jkth4ekYcB3gYOAnYEjJO1c0+wzwDMR8VfAt4FvttqvmXVWFXseewBLI+LxiHgZuAqYXtNmOjA3TV8L7CtJFfRtZh1SRXhMBJaVnvek1+q2iYj1wBpgXAV9m1mHDK9gGfX2ICKjDZJmAjMBNmGz1iszs0FTxZ5HDzC59HwSsLxRG0nDgS2B1bULiog5ETElIqaMYGQFpZnZYKkiPO4GdpT0dkkbAzOA+TVt5gNHp+nDgVsiYoM9DzMbOlo+bImI9ZJOAH4GDAMuiYjFks4AFkXEfOBi4HJJSyn2OGa02q+ZdVYV5zyIiAXAgprXTitNvwh8rIq+zKw7+A5TM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLA4PM8vi8DCzLJWEh6QDJT0iaamkWXXmHyPpaUn3pcexVfRrZp0zvNUFSBoGfBfYH+gB7pY0PyIermk6LyJOaLU/M+sOVex57AEsjYjHI+Jl4CpgegXLNbMu1vKeBzARWFZ63gPsWafdYZI+CPwO+GJELKttIGkmMBNg0sRhXH3XHRWU98a05UabdrqErvfecz7b6RLe0KrY81Cd16Lm+Q3A9hHxHuBmYG69BUXEnIiYEhFTxo/zuVyzblbFFtoDTC49nwQsLzeIiFUR8VJ6ehGwewX9mlkHVREedwM7Snq7pI2BGcD8cgNJE0pPDwWWVNCvmXVQy+c8ImK9pBOAnwHDgEsiYrGkM4BFETEf+LykQ4H1wGrgmFb7NbPOquKEKRGxAFhQ89pppemTgZOr6MvMuoPPSppZFoeHmWVxeJhZFoeHmWVxeJhZFoeHmWVxeJhZFoeHmWVxeJhZFoeHmWVxeJhZFoeHmWVxeJhZFoeHmWVxeJhZFoeHmWVxeJhZFoeHmWVxeJhZFoeHmWVxeJhZFoeHmWVxeJhZFoeHmWVxeJhZFoeHmWVxeJhZlkrCQ9Ilkp6S9FCD+ZJ0vqSlkh6QtFsV/ZpZ51S153EpcGAf8w8CdkyPmcCFFfVrZh1SSXhExO3A6j6aTAcui8JCYCtJE6ro28w6o13nPCYCy0rPe9JrryNppqRFkhatXPVam0ozsxztCg/VeS02eCFiTkRMiYgp48f5XK5ZN2vXFtoDTC49nwQsb1PfZjYI2hUe84Gj0lWXvYA1EbGiTX2b2SAYXsVCJF0JTAPGS+oBTgdGAETEbGABcDCwFHge+HQV/ZpZ51QSHhFxRD/zA/hcFX2ZWXfwWUkzy+LwMLMsDg8zy+LwMLMsDg8zy+LwMLMsDg8zy+LwMLMsDg8zy+LwMLMsDg8zy+LwMLMsDg8zy+LwMLMsDg8zy+LwMLMsDg8zy+LwMLMsDg8zy+LwMLMsDg8zy+LwMLMsDg8zy+LwMLMsDg8zy+LwMLMsDg8zy1JJeEi6RNJTkh5qMH+apDWS7kuP06ro18w6p5L/6Bq4FLgAuKyPNr+KiEMq6s/MOqySPY+IuB1YXcWyzGxoqGrPoxl7S7ofWA58KSIW1zaQNBOYCTBszBh2u+aLbSxvaNl40p86XULXm/uF8zpdQtfb+9v5723XCdN7ge0i4r3AvwPX12sUEXMiYkpETBk2avM2lWZmOdoSHhGxNiLWpekFwAhJ49vRt5kNjraEh6RtJClN75H6XdWOvs1scFRyzkPSlcA0YLykHuB0YARARMwGDgeOl7QeeAGYERFRRd9m1hmVhEdEHNHP/AsoLuWa2RuE7zA1sywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywODzPL4vAwsywth4ekyZJulbRE0mJJX6jTRpLOl7RU0gOSdmu1XzPrrCr+o+v1wEkRca+k0cA9km6KiIdLbQ4CdkyPPYEL008zG6Ja3vOIiBURcW+afg5YAkysaTYduCwKC4GtJE1otW8z65xKz3lI2h54H3BnzayJwLLS8x42DBgzG0IqCw9Jo4DrgBMjYm3t7DpviTrLmClpkaRFr677U1WlmdkgqCQ8JI2gCI4rIuKHdZr0AJNLzycBy2sbRcSciJgSEVOGjdq8itLMbJBUcbVFwMXAkog4t0Gz+cBR6arLXsCaiFjRat9m1jlVXG3ZBzgSeFDSfem1U4BtASJiNrAAOBhYCjwPfLqCfs2sg1oOj4j4NfXPaZTbBPC5Vvsys+7hO0zNLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yyODzMLIvDw8yytBwekiZLulXSEkmLJX2hTptpktZIui89Tmu1XzPrrOEVLGM9cFJE3CtpNHCPpJsi4uGadr+KiEMq6M/MukDLex4RsSIi7k3TzwFLgImtLtfMupsiorqFSdsDtwO7RsTa0uvTgOuAHmA58KWIWFzn/TOBmenprsBDlRVXjfHAyk4XUeJ6+tZt9UD31fSuiBid88bKwkPSKOCXwJkR8cOaeVsAr0XEOkkHA+dFxI79LG9RREyppLiKdFtNrqdv3VYPdF9NrdRTydUWSSMo9iyuqA0OgIhYGxHr0vQCYISk8VX0bWadUcXVFgEXA0si4twGbbZJ7ZC0R+p3Vat9m1nnVHG1ZR/gSOBBSfel104BtgWIiNnA4cDxktYDLwAzov/jpTkV1Fa1bqvJ9fSt2+qB7qspu55KT5ia2ZuH7zA1sywODzPL0jXhIWmspJskPZp+jmnQ7tXSbe7zB6GOAyU9ImmppFl15o+UNC/NvzPd2zKomqjpGElPl8bl2EGs5RJJT0mqew+OCuenWh+QtNtg1TKAmtr29Ygmv67R1jEatK+QRERXPIBzgFlpehbwzQbt1g1iDcOAx4AdgI2B+4Gda9p8FpidpmcA8wZ5XJqp6RjggjZ9Th8EdgMeajD/YOAngIC9gDu7oKZpwI1tGp8JwG5pejTwuzqfV1vHqMmaBjxGXbPnAUwH5qbpucBHOlDDHsDSiHg8Il4Grkp1lZXrvBbYt/cydAdrapuIuB1Y3UeT6cBlUVgIbCVpQodrapto7usabR2jJmsasG4Kj7dGxAooflngLQ3abSJpkaSFkqoOmInAstLzHjYc5D+3iYj1wBpgXMV1DLQmgMPSLvC1kiYPYj39abbedttb0v2SfiJpl3Z0mA5p3wfcWTOrY2PUR00wwDGq4j6Ppkm6GdimzqxTB7CYbSNiuaQdgFskPRgRj1VTIfX2IGqvZTfTpkrN9HcDcGVEvCTpOIo9o38YxJr60u7xaca9wHbxl69HXA/0+fWIVqWva1wHnBil73n1zq7zlkEfo35qGvAYtXXPIyL2i4hd6zx+BDzZu+uWfj7VYBnL08/HgdsoUrQqPUD5r/Ykii/y1W0jaTiwJYO7y9xvTRGxKiJeSk8vAnYfxHr608wYtlW0+esR/X1dgw6M0WB8haSbDlvmA0en6aOBH9U2kDRG0sg0PZ7i7tbafzekFXcDO0p6u6SNKU6I1l7RKdd5OHBLpDNOg6TfmmqOlw+lOKbtlPnAUemKwl7Amt7D0U5p59cjUj99fl2DNo9RMzVljVE7zkA3eUZ4HPAL4NH0c2x6fQrw/TQ9FXiQ4orDg8BnBqGOgynORj8GnJpeOwM4NE1vAlwDLAXuAnZow9j0V9NZwOI0LrcC7x7EWq4EVgCvUPwF/QxwHHBcmi/gu6nWB4EpbRif/mo6oTQ+C4Gpg1jL+ykOQR4A7kuPgzs5Rk3WNOAx8u3pZpalmw5bzGwIcXiYWRaHh5llcXiYWRaHh5llcXiYWRaHh5ll+X+iieLond1zFAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "for l in mm.layers:\n",
    "    if type(l).__name__ == \"Conv2d\":\n",
    "        for f in range(l.filters):\n",
    "            for c in range(l.weights.shape[2]):\n",
    "                plt.imshow(l.weights[:, :, c, f])\n",
    "                plt.title(f\"Layer: {l.name} Filter: {f} Channel: {c}\")\n",
    "                plt.show()\n",
    "    if type(l).__name__ == \"FFL\":\n",
    "        plt.imshow(l.weights)\n",
    "        plt.title(l.name)\n",
    "        plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-05-22T06:57:44.803898Z",
     "start_time": "2020-05-22T06:57:44.715951Z"
    },
    "code_folding": [
     0,
     36,
     51,
     52,
     130,
     131
    ]
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(3072,)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "class Dropout:\n",
    "    def __init__(self, prob = 0.5):\n",
    "        self.input_shape=None\n",
    "        self.output_shape = None\n",
    "        self.input_data= None\n",
    "        self.output = None\n",
    "        self.isbias = False\n",
    "        self.activation = None\n",
    "        self.parameters = 0\n",
    "        self.delta = 0\n",
    "        self.weights = 0\n",
    "        self.bias = 0\n",
    "        self.prob = prob\n",
    "        self.dw = 0\n",
    "        self.db = 0\n",
    "        \n",
    "    def set_output_shape(self):\n",
    "        #print(self.input_shape)\n",
    "        self.output_shape = self.input_shape\n",
    "        self.weights = 0\n",
    "    def apply_activation(self, x):\n",
    "        self.input_data = x\n",
    "        #print(x.sum(axis=2))\n",
    "        flat = np.array(self.input_data).flatten()\n",
    "        random_indices = np.random.randint(0, len(flat), int(self.prob * len(flat)))\n",
    "        flat[random_indices] = 0\n",
    "        self.output = flat.reshape(x.shape)\n",
    "        return self.output\n",
    "    def activation_dfn(self, x):\n",
    "        return x\n",
    "           \n",
    "x = np.arange(0, 100).reshape(10, 10)\n",
    "dp = Dropout()\n",
    "dp.apply_activation(x)\n",
    "\n",
    "yy = np.arange(1, 11)\n",
    "def one_hot_encoding(lbl, classes):\n",
    "    encoded = np.zeros((len(lbl), classes))\n",
    "    c = list(set(lbl))\n",
    "    if len(c) != classes:\n",
    "        raise ValueError(\"Number of classes is not equal to unique labels.\")\n",
    "    for i in range(len(yy)):\n",
    "        for j in range(len(c)):\n",
    "            if c[j] == lbl[i]:\n",
    "                encoded[i, j] = 1\n",
    "    return encoded\n",
    "#yy = l.copy()\n",
    "#c = len(set(l))\n",
    "#ohe = one_hot_encoding(yy, c)      \n",
    "#print(ohe[0])\n",
    "\n",
    "class Pool2d:\n",
    "    def __init__(self, kernel_size = (2, 2), stride=(1, 1), kind=\"max\"):\n",
    "        self.input_shape=None\n",
    "        self.output_shape = None\n",
    "        self.input_data= None\n",
    "        self.output = None\n",
    "        self.isbias = False\n",
    "        self.activation = None\n",
    "        self.parameters = 0\n",
    "        self.delta = 0\n",
    "        self.weights = 0\n",
    "        self.bias = 0\n",
    "        \n",
    "        self.kernel_size = kernel_size\n",
    "        if type(stride) == int:\n",
    "                 stride = (stride, stride)\n",
    "\n",
    "        self.stride = stride\n",
    "        self.pools = ['max', \"average\", 'min']\n",
    "        if kind not in self.pools:\n",
    "            raise ValueError(\"Pool kind not understoood.\")\n",
    "            \n",
    "        self.kind = kind\n",
    "    \n",
    "    def set_output_shape(self):\n",
    "        shape = self.input_shape\n",
    "        stride = self.stride\n",
    "        kshape = self.kernel_size\n",
    "        rstep = stride[0]+kshape[0]-1\n",
    "        cstep = stride[1]+kshape[1]-1\n",
    "        rv = rstep * len(range(kshape[0], shape[0]+1, rstep))\n",
    "        cv = cstep * len(range(kshape[1], shape[1]+1, cstep))\n",
    "        self.output_shape = (int(rv/rstep), int(cv/cstep), shape[2])\n",
    "#         print(self.output_shape)\n",
    "        self.weights = 0\n",
    "    def apply_activation(self, image):\n",
    "        stride = self.stride\n",
    "        kshape = self.kernel_size\n",
    "        shape = image.shape\n",
    "        self.input_shape = shape\n",
    "        self.set_output_shape()\n",
    "        #print(self.output_shape, shape)\n",
    "        \n",
    "        \n",
    "        rstep = stride[0]+kshape[0]-1\n",
    "        cstep = stride[1]+kshape[1]-1\n",
    "        self.out = np.zeros((self.output_shape))\n",
    "        for nc in range(shape[2]):\n",
    "            cimg = []\n",
    "            rv = 0\n",
    "            for r in range(kshape[0], shape[0]+1, rstep):\n",
    "                cv = 0\n",
    "                for c in range(kshape[1], shape[1]+1, cstep):\n",
    "                    chunk = image[rv:r, cv:c, nc]\n",
    "                    if len(chunk) > 0:\n",
    "                        \n",
    "                        if self.kind == \"max\":\n",
    "                            #print(chunk)\n",
    "                            chunk = np.max(chunk)\n",
    "                        if self.kind == \"min\":\n",
    "                            chunk = np.min(chunk)\n",
    "                        if self.kind == \"average\":\n",
    "                            chunk = np.mean(chunk)\n",
    "\n",
    "                        cimg.append(chunk)\n",
    "                    else:\n",
    "                        cv-=cstep\n",
    "                        #rv-=rstep\n",
    "                    cv+=cstep\n",
    "                rv+=rstep\n",
    "            cimg = np.array(cimg).reshape(int(rv/rstep), int(cv/cstep))\n",
    "            #print(cimg.shape)\n",
    "            self.out[:,:,nc] = cimg\n",
    "        return self.out\n",
    "pool = Pool2d(kernel_size=(7, 7), kind=\"max\")\n",
    "test = np.random.randint(1, 100, (32, 32, 3))\n",
    "o = pool.apply_activation(test)\n",
    "o.shape\n",
    "\n",
    "class Flatten:\n",
    "    def __init__(self, input_shape=None):\n",
    "        self.input_shape=None\n",
    "        self.output_shape = None\n",
    "        self.input_data= None\n",
    "        self.output = None\n",
    "        self.isbias = False\n",
    "        self.activation = None\n",
    "        self.parameters = 0\n",
    "        self.delta = 0\n",
    "        self.weights = 0\n",
    "        self.bias = 0\n",
    "        \n",
    "    def set_output_shape(self):\n",
    "        #print(self.input_shape)\n",
    "        self.output_shape = (self.input_shape[0] * self.input_shape[1] * self.input_shape[2])\n",
    "        self.weights = 0\n",
    "    def apply_activation(self, x):\n",
    "        self.input_data = x\n",
    "        #print(x.sum(axis=2))\n",
    "        self.output = np.array(self.input_data).flatten()\n",
    "        return self.output\n",
    "    def activation_dfn(self, x):\n",
    "        return x\n",
    "#x = np.array([[1, 1, 1], [1, 0, 1], [0, 1, 1], [0, 0, 1]])\n",
    "#f = Flatten()\n",
    "#f.apply_activation(test).shape\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-05-22T06:26:58.495583Z",
     "start_time": "2020-05-22T06:26:43.017Z"
    }
   },
   "outputs": [],
   "source": [
    "yy = np.arange(1, 11)\n",
    "def one_hot_encoding(lbl, classes):\n",
    "    encoded = np.zeros((len(lbl), classes))\n",
    "    c = list(set(lbl))\n",
    "    if len(c) != classes:\n",
    "        raise ValueError(\"Number of classes is not equal to unique labels.\")\n",
    "    for i in range(len(yy)):\n",
    "        for j in range(len(c)):\n",
    "            if c[j] == lbl[i]:\n",
    "                encoded[i, j] = 1\n",
    "    return encoded\n",
    "yy = l.copy()\n",
    "c = len(set(l))\n",
    "ohe = one_hot_encoding(yy, c)      \n",
    "print(ohe[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-05-22T06:26:58.497583Z",
     "start_time": "2020-05-22T06:26:43.126Z"
    },
    "code_folding": [
     0
    ]
   },
   "outputs": [],
   "source": [
    "class Pool2d:\n",
    "    def __init__(self, kernel_size = (2, 2), stride=(1, 1), kind=\"max\"):\n",
    "        self.input_shape=None\n",
    "        self.output_shape = None\n",
    "        self.input_data= None\n",
    "        self.output = None\n",
    "        self.isbias = False\n",
    "        self.activation = None\n",
    "        self.parameters = 0\n",
    "        self.delta = 0\n",
    "        self.weights = 0\n",
    "        self.bias = 0\n",
    "        \n",
    "        self.kernel_size = kernel_size\n",
    "        if type(stride) == int:\n",
    "                 stride = (stride, stride)\n",
    "\n",
    "        self.stride = stride\n",
    "        self.pools = ['max', \"average\", 'min']\n",
    "        if kind not in self.pools:\n",
    "            raise ValueError(\"Pool kind not understoood.\")\n",
    "            \n",
    "        self.kind = kind\n",
    "    \n",
    "    def set_output_shape(self):\n",
    "        shape = self.input_shape\n",
    "        stride = self.stride\n",
    "        kshape = self.kernel_size\n",
    "        rstep = stride[0]+kshape[0]-1\n",
    "        cstep = stride[1]+kshape[1]-1\n",
    "        rv = rstep * len(range(kshape[0], shape[0]+1, rstep))\n",
    "        cv = cstep * len(range(kshape[1], shape[1]+1, cstep))\n",
    "        self.output_shape = (int(rv/rstep), int(cv/cstep), shape[2])\n",
    "#         print(self.output_shape)\n",
    "        self.weights = 0\n",
    "    def apply_activation(self, image):\n",
    "        stride = self.stride\n",
    "        kshape = self.kernel_size\n",
    "        shape = image.shape\n",
    "        self.input_shape = shape\n",
    "        self.set_output_shape()\n",
    "        #print(self.output_shape, shape)\n",
    "        \n",
    "        \n",
    "        rstep = stride[0]+kshape[0]-1\n",
    "        cstep = stride[1]+kshape[1]-1\n",
    "        self.out = np.zeros((self.output_shape))\n",
    "        for nc in range(shape[2]):\n",
    "            cimg = []\n",
    "            rv = 0\n",
    "            for r in range(kshape[0], shape[0]+1, rstep):\n",
    "                cv = 0\n",
    "                for c in range(kshape[1], shape[1]+1, cstep):\n",
    "                    chunk = image[rv:r, cv:c, nc]\n",
    "                    if len(chunk) > 0:\n",
    "                        \n",
    "                        if self.kind == \"max\":\n",
    "                            #print(chunk)\n",
    "                            chunk = np.max(chunk)\n",
    "                        if self.kind == \"min\":\n",
    "                            chunk = np.min(chunk)\n",
    "                        if self.kind == \"average\":\n",
    "                            chunk = np.mean(chunk)\n",
    "\n",
    "                        cimg.append(chunk)\n",
    "                    else:\n",
    "                        cv-=cstep\n",
    "                        #rv-=rstep\n",
    "                    cv+=cstep\n",
    "                rv+=rstep\n",
    "            cimg = np.array(cimg).reshape(int(rv/rstep), int(cv/cstep))\n",
    "            #print(cimg.shape)\n",
    "            self.out[:,:,nc] = cimg\n",
    "        return self.out\n",
    "pool = Pool2d(kernel_size=(7, 7), kind=\"max\")\n",
    "test = np.random.randint(1, 100, (32, 32, 3))\n",
    "o = pool.apply_activation(test)\n",
    "o.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-01T13:45:15.567240Z",
     "start_time": "2020-06-01T13:45:15.148483Z"
    },
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x296ab953e88>"
      ]
     },
     "execution_count": 77,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAPsAAAD4CAYAAAAq5pAIAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAPEklEQVR4nO3dXYxc5X3H8d9vxzaGtTF+4cU1TnCJmwaSYqrFJQW1RCkIHFUmF6nwReSqKM5FkIiaSiX0Iki9oVUDiqoUyRQUg1JSpAThtjSJa6GiVFXEgoxf4qZ2YcFrb73GDni9Zu3d9b8Xe6jWsOeZZc68rZ/vR1rN7PnPmfP38f7mzO4z5zyOCAG48PV0ugEA7UHYgUwQdiAThB3IBGEHMjGvnRurLeqNecuXlj/gnNvXDDBHOTGANv6rE5ocHZ0xSJXCbvtOSd+RVJP09xHxcOrx85Yv1VXfvL+03jPGGw2gntqZ8oPi4N8+WlprOF22a5K+K+kuSddJ2mT7ukafD0BrVTmUrpd0MCJej4izkn4gaWNz2gLQbFXCvkrSoWnfDxbLzmN7i+1+2/2Tp0YrbA5AFVXCPtMvDh/600FEbI2Ivojoqy3qrbA5AFVUCfugpNXTvr9a0pFq7QBolSphf1nSWttrbC+QdI+k7c1pC0CzNTz0FhETtu+T9BNNDb09GRH7kitZUo2z7IAqPJEoJuJVaZw9Il6Q9EKV5wDQHnyKBcgEYQcyQdiBTBB2IBOEHcgEYQcyQdiBTBB2IBOEHcgEYQcyQdiBTBB2IBOEHcgEYQcyQdiBTBB2IBOEHcgEYQcyQdiBTBB2IBOEHcgEYQcyQdiBTBB2IBOEHcgEYQcyQdiBTBB2IBOEHcgEYQcyUWnKZtsDkkYkTUqaiIi+ZjQFoPkqhb3wuYh4uwnPA6CFeBsPZKJq2EPST22/YnvLTA+wvcV2v+3+yZHRipsD0Kiqb+NviYgjtq+QtMP2f0XES9MfEBFbJW2VpIuuuToqbg9Agyod2SPiSHE7LOk5Seub0RSA5ms47LZ7bS9+/76kOyTtbVZjAJqrytv4KyU9Z/v95/mHiPhxU7rqgAUn0697C94pr51Zlv7t5OzSc8l6XDKZrLuWfv4Yq5WvO+H0c59N13HhaDjsEfG6pBua2AuAFmLoDcgEYQcyQdiBTBB2IBOEHchEM06EuSAsOZgeHusdGi+tDX32oma3c76e9NCbF5YP3cXZ9Ou5z5YP23W7mJfeLz3Lz5bXaunhzvETC9PPPTb3jpNzr2MADSHsQCYIO5AJwg5kgrADmSDsQCYIO5AJxtkLC06lx9kXHE1dUqvaOHvPyTr/DXXOQo3UOPzC9L+rnp7x9MbPzW/84kN1x8nrbFt1Tt+9fNnJ0tpNl7+VXPffBj6ZrJ95a1Gy3o04sgOZIOxAJgg7kAnCDmSCsAOZIOxAJgg7kAnG2QtjS9LndcfaJaW1iUV1zjevM1684N0655xPJMvJl+z3Vtbprc4weW2s4jh7ore4qM41BN6Yn6zXxtLbvu53j5bW7r/8xeS6+95ZmawPMM4OoFsRdiAThB3IBGEHMkHYgUwQdiAThB3IBOPshXc/ka7XEtMin5tf7Zzxelzn6XvKL4/e1dc3r42kP9uwfM+ZZH2iN73+tZccK60t60nvl9Pj6TH+uajuT4LtJ20P2947bdky2ztsHyhul7a2TQBVzeZl/3uS7vzAsgck7YyItZJ2Ft8D6GJ1wx4RL0k68YHFGyVtK+5vk3R3k/sC0GSN/kJ3ZUQMSVJxe0XZA21vsd1vu39yJHUdNwCt1PK/3kTE1ojoi4i+2uLeVm8OQIlGw37U9kpJKm6Hm9cSgFZoNOzbJW0u7m+W9Hxz2gHQKnXH2W0/I+k2SStsD0r6lqSHJT1r+15Jb0n6UiubbIeJ3vS50fXqVZxdkh5I92T6nPLaWDO7OV+V68JL6XP5l+5Pr7tw10CyPnRP+truv7+ofAP/NPqx5LpHj5Vfv0Cqeyn/rlQ37BGxqaT0+Sb3AqCFuvfjVQCairADmSDsQCYIO5AJwg5kglNc54Co1RsWrPDcdcaQJhdWG3q7+Fj5Bi7/z/JTUCVJl12aLPuO48n6DQvKz/39q0N96W3/akG6PgdxZAcyQdiBTBB2IBOEHcgEYQcyQdiBTBB2IBOMs6OSee+lB+qX7xsvLw6lr3ly6CvXJ+vf/fTfJev/PFo+7fLeN38tua4nk+U5iSM7kAnCDmSCsAOZIOxAJgg7kAnCDmSCsAOZYJwdlSwcTo+z9+46XFqb+I305Zw//oU3kvXLetLX0P7LwVvLiyMX3pTM9XBkBzJB2IFMEHYgE4QdyARhBzJB2IFMEHYgE4yzI8np2aS19GDifHVJE0eGSmuH/uSa5Lr/suaxZP25kd9K1g8eKD+fvefMXJx0uZq6R3bbT9oetr132rKHbB+2vav42tDaNgFUNZu38d+TdOcMyx+NiHXF1wvNbQtAs9UNe0S8JOlEG3oB0EJV/kB3n+3dxdv8pWUPsr3Fdr/t/smR0QqbA1BFo2F/TNK1ktZJGpL07bIHRsTWiOiLiL7a4gozEAKopKGwR8TRiJiMiHOSHpe0vrltAWi2hsJue/qYxhcl7S17LIDuUHec3fYzkm6TtML2oKRvSbrN9jpJIWlA0ldb2CM66KLj6ePBJb88mqzHZz5ZWvvN2w801NP7nhlIz7Hec5rPjE1XN+wRsWmGxU+0oBcALcRLH5AJwg5kgrADmSDsQCYIO5AJTnHN3PxT6VM9V+ydSNZ9Jn2K68Cm8tNMn179VHLdZ0/emKy/feiyZJ0j2fnYH0AmCDuQCcIOZIKwA5kg7EAmCDuQCcIOZIJx9gtcz0R6HH1JnbNMF+3532T95E2rkvXPfmF3aW2hJ5PrPjuQHmfveY9j1UfB3gIyQdiBTBB2IBOEHcgEYQcyQdiBTBB2IBOMs18AHOW1RQPpcfZlr6Wn8Zu4Kn3O+OAfps93f2LlT0prjx+/NbnuiUHOV28m9heQCcIOZIKwA5kg7EAmCDuQCcIOZIKwA5lgnP0C0DNePpa+Yvfp5LoeOJKsv/ln1yfr37z5+WT95bGPldb+deC65Lqcr95cdfem7dW2X7S93/Y+2/cXy5fZ3mH7QHG7tPXtAmjUbF46JyR9IyI+JelmSV+zfZ2kByTtjIi1knYW3wPoUnXDHhFDEfFqcX9E0n5JqyRtlLSteNg2SXe3qkkA1X2kX4psXyPpRkk/l3RlRAxJUy8Ikq4oWWeL7X7b/ZMjo9W6BdCwWYfd9iJJP5T09Yg4Odv1ImJrRPRFRF9tcW8jPQJoglmF3fZ8TQX9+xHxo2LxUdsri/pKScOtaRFAM9QderNtSU9I2h8Rj0wrbZe0WdLDxW16DAYN6zmbPk310sTloOfvfSO57snbP5Ws/8Fdrybrq+anT5F9ZOCO0trpw4uS66b/1fioZjPOfoukL0vaY3tXsexBTYX8Wdv3SnpL0pda0yKAZqgb9oj4mcpfZD/f3HYAtAofUQIyQdiBTBB2IBOEHcgEYQcywSmuc8DCt9MjziteK/9Ao5elT0Y8vCE9bfLTV+xM1h899rlk/fXXryytpU7NRfNxZAcyQdiBTBB2IBOEHcgEYQcyQdiBTBB2IBOMs3eBeafT483LfzGerPe8UX456HdvW5tc909v/nGyvmZ++pzzfx/8RLLec6qWrKN9OLIDmSDsQCYIO5AJwg5kgrADmSDsQCYIO5AJxtm7wMLj6XH23t3paZVjefk568evT49z33Dxm8n6qXNjyfro6MJkHd2DIzuQCcIOZIKwA5kg7EAmCDuQCcIOZIKwA5mYzfzsqyU9JekqSeckbY2I79h+SNJXJB0rHvpgRLzQqkYvZBcfO5esTwweTtbP3HVTaW3JzcPJdWtKb/uxd65P1iffS4/jczTpHrP5UM2EpG9ExKu2F0t6xfaOovZoRPxN69oD0CyzmZ99SNJQcX/E9n5Jq1rdGIDm+kjvsmxfI+lGST8vFt1ne7ftJ23P+JlN21ts99vunxwZrdQsgMbNOuy2F0n6oaSvR8RJSY9JulbSOk0d+b8903oRsTUi+iKir7a4twktA2jErMJue76mgv79iPiRJEXE0YiYjIhzkh6XtL51bQKoqm7YbVvSE5L2R8Qj05avnPawL0ra2/z2ADTLbP4af4ukL0vaY3tXsexBSZtsr5MUkgYkfbUlHWZgvLfO1MXrP5Msv7tmfmntkp700No/nvidZP0/htYk6x7jUtFzxWz+Gv8zSTP9NDKmDswhfOYByARhBzJB2IFMEHYgE4QdyARhBzLR3ktJh6TJOmPKGRpbnt4nx29IT5s8dkV5rWc8/V/84qH0lM6jb12arHsyWUYLROq/NPGjxJEdyARhBzJB2IFMEHYgE4QdyARhBzJB2IFMOCLatzH7mKTpcwSvkPR22xr4aLq1t27tS6K3RjWzt49HxOUzFdoa9g9t3O6PiL6ONZDQrb11a18SvTWqXb3xNh7IBGEHMtHpsG/t8PZTurW3bu1LordGtaW3jv7ODqB9On1kB9AmhB3IREfCbvtO27+0fdD2A53ooYztAdt7bO+y3d/hXp60PWx777Rly2zvsH2guJ1xjr0O9faQ7cPFvttle0OHeltt+0Xb+23vs31/sbyj+y7RV1v2W9t/Z7ddk/Tfkm6XNCjpZUmbIuIXbW2khO0BSX0R0fEPYNj+PUmnJD0VEZ8ulv21pBMR8XDxQrk0Iv68S3p7SNKpTk/jXcxWtHL6NOOS7pb0x+rgvkv09Udqw37rxJF9vaSDEfF6RJyV9ANJGzvQR9eLiJcknfjA4o2SthX3t2nqh6XtSnrrChExFBGvFvdHJL0/zXhH912ir7boRNhXSTo07ftBddd87yHpp7Zfsb2l083M4MqIGJKmfngkJS5K1RF1p/Fupw9MM941+66R6c+r6kTYZ7pKVjeN/90SEb8t6S5JXyvermJ2ZjWNd7vMMM14V2h0+vOqOhH2QUmrp31/taQjHehjRhFxpLgdlvScum8q6qPvz6Bb3A53uJ//103TeM80zbi6YN91cvrzToT9ZUlrba+xvUDSPZK2d6CPD7HdW/zhRLZ7Jd2h7puKerukzcX9zZKe72Av5+mWabzLphlXh/ddx6c/j4i2f0naoKm/yP+PpL/oRA8lff26pNeKr32d7k3SM5p6WzeuqXdE90paLmmnpAPF7bIu6u1pSXsk7dZUsFZ2qLdbNfWr4W5Ju4qvDZ3ed4m+2rLf+LgskAk+QQdkgrADmSDsQCYIO5AJwg5kgrADmSDsQCb+Dx2SbtWO7mAbAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#img = np.random.randint(0, 255, (10, 10, 1))\n",
    "img = xt[0]\n",
    "conv = Conv2d()\n",
    "conv.input=img\n",
    "conv.weights = np.array([[1, 0, -1], [1, 0, -1], [1, 0, -1]]).reshape(3, 3, 1, 1)\n",
    "conv.biases = np.zeros(1)\n",
    "conv.out = np.zeros((28, 28, 1))\n",
    "cout = conv.apply_activation(img)\n",
    "plt.imshow(cout.reshape(28, 28))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-05-22T06:26:58.507575Z",
     "start_time": "2020-05-22T06:26:43.416Z"
    },
    "code_folding": [
     0
    ]
   },
   "outputs": [],
   "source": [
    "class Flatten:\n",
    "    def __init__(self, input_shape=None):\n",
    "        self.input_shape=None\n",
    "        self.output_shape = None\n",
    "        self.input_data= None\n",
    "        self.output = None\n",
    "        self.isbias = False\n",
    "        self.activation = None\n",
    "        self.parameters = 0\n",
    "        self.delta = 0\n",
    "        self.weights = 0\n",
    "        self.bias = 0\n",
    "        \n",
    "    def set_output_shape(self):\n",
    "        #print(self.input_shape)\n",
    "        self.output_shape = (self.input_shape[0] * self.input_shape[1] * self.input_shape[2])\n",
    "        self.weights = 0\n",
    "    def apply_activation(self, x):\n",
    "        self.input_data = x\n",
    "        #print(x.sum(axis=2))\n",
    "        self.output = np.array(self.input_data).flatten()\n",
    "        return self.output\n",
    "    def activation_dfn(self, x):\n",
    "        return x\n",
    "#x = np.array([[1, 1, 1], [1, 0, 1], [0, 1, 1], [0, 0, 1]])\n",
    "f = Flatten()\n",
    "f.apply_activation(test).shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-11T02:55:05.335147Z",
     "start_time": "2020-06-11T02:55:05.301167Z"
    },
    "code_folding": [
     0
    ]
   },
   "outputs": [],
   "source": [
    "class Unsample:\n",
    "    def __init__(self, kernel_size = (2, 2)):\n",
    "        self.input_shape=None\n",
    "        self.output_shape = None\n",
    "        self.input_data= None\n",
    "        self.output = None\n",
    "        self.isbias = False\n",
    "        self.activation = None\n",
    "        self.parameters = 0\n",
    "        self.delta = 0\n",
    "        self.weights = 0\n",
    "        self.bias = 0\n",
    "        self.delta_weights = 0\n",
    "        self.delta_biases = 0        \n",
    "        self.kernel_size = kernel_size\n",
    "        self.stride = self.kernel_size\n",
    "        \n",
    "    \n",
    "    def set_output_shape(self):\n",
    "        shape = self.input_shape\n",
    "        self.output_shape = (shape[0] * self.kernel_size[0], shape[1] * self.kernel_size[1], shape[2])\n",
    "        self.weights = 0\n",
    "    def apply_activation(self, image):\n",
    "        stride = self.stride\n",
    "        kshape = self.kernel_size\n",
    "        self.input_shape = image.shape\n",
    "        self.set_output_shape()\n",
    "        \n",
    "        rstep = stride[0]\n",
    "        cstep = stride[1]\n",
    "        self.out = np.zeros(self.output_shape)\n",
    "        shape = self.output_shape\n",
    "        \n",
    "        \n",
    "        for nc in range(shape[2]):\n",
    "            cimg = []\n",
    "            rv = 0\n",
    "            i = 0\n",
    "            for r in range(kshape[0], shape[0]+1, rstep):\n",
    "                cv = 0\n",
    "                j = 0\n",
    "                for c in range(kshape[1], shape[1]+1, cstep):\n",
    "                    self.out[rv:r, cv:c] = image[i, j]\n",
    "                    j+=1\n",
    "                    cv+=cstep\n",
    "                rv+=rstep\n",
    "                i+=1\n",
    "                \n",
    "        return self.out\n",
    "    def backpropagate(self, nx_layer):\n",
    "        \"\"\"\n",
    "            Gradients are passed through index of largest value .\n",
    "        \"\"\"\n",
    "        layer = self\n",
    "        stride = layer.stride\n",
    "        kshape = layer.kernel_size\n",
    "        image = layer.input\n",
    "        shape = image.shape\n",
    "        layer.delta = np.zeros(shape)\n",
    "        #self.input_shape = shape\n",
    "        #self.set_output_shape()\n",
    "        #print(self.output_shape)\n",
    "        \n",
    "        cimg = []\n",
    "        rstep = stride[0]\n",
    "        cstep = stride[1]\n",
    "        \n",
    "        for f in range(shape[2]):\n",
    "            i = 0\n",
    "            rv = 0\n",
    "            for r in range(kshape[0], shape[0]+1, rstep):\n",
    "                cv = 0\n",
    "                j = 0\n",
    "                for c in range(kshape[1], shape[1]+1, cstep):\n",
    "                    chunk = image[rv:r, cv:c, f]\n",
    "                    dout = nx_layer.delta[i, j, f]\n",
    "                    \n",
    "                    if layer.kind == \"max\":\n",
    "                        p = np.max(chunk)\n",
    "                        #print(p, chunk)\n",
    "                        index = np.argwhere(chunk == p)[0]\n",
    "                        #print(layer.delta[rv+index[0], cv+index[1], f].shape, dout.shape)\n",
    "                        layer.delta[rv+index[0], cv+index[1], f] = dout\n",
    "                        #print(index)\n",
    "                    if layer.kind == \"min\":\n",
    "                        p = np.min(chunk)\n",
    "                        index = np.argwhere(chunk == p)[0]\n",
    "                        layer.delta[rv+index[0], cv+index[1], f] = dout\n",
    "                    if layer.kind == \"average\":\n",
    "                        p = np.mean(chunk)\n",
    "                        layer.delta[rv:r, cv:c, f] = dout\n",
    "\n",
    "\n",
    "                    j+=1\n",
    "                    cv+=cstep\n",
    "                rv+=rstep\n",
    "                i+=1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-11T02:55:06.689161Z",
     "start_time": "2020-06-11T02:55:06.675162Z"
    }
   },
   "outputs": [],
   "source": [
    "us = Unsample(kernel_size=(1, 3))\n",
    "img = us.apply_activation(x_train[0].reshape(28, 28, 1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-11T02:52:33.592103Z",
     "start_time": "2020-06-11T02:52:33.321254Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x2ea255f3608>"
      ]
     },
     "execution_count": 75,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(img.reshape(28, 28*3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-11T02:42:01.228090Z",
     "start_time": "2020-06-11T02:42:00.868299Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x2ea23b2ed48>"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(x_train[0].reshape(28, 28))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-11T02:42:23.144713Z",
     "start_time": "2020-06-11T02:42:23.135718Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.0"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "img.max()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-11T02:47:14.086098Z",
     "start_time": "2020-06-11T02:47:14.078106Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "255"
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_train[0].reshape(28, 28, 1).max()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  },
  "latex_envs": {
   "LaTeX_envs_menu_present": true,
   "autoclose": false,
   "autocomplete": true,
   "bibliofile": "biblio.bib",
   "cite_by": "apalike",
   "current_citInitial": 1,
   "eqLabelWithNumbers": true,
   "eqNumInitial": 1,
   "hotkeys": {
    "equation": "Ctrl-E",
    "itemize": "Ctrl-I"
   },
   "labels_anchors": false,
   "latex_user_defs": false,
   "report_style_numbering": false,
   "user_envs_cfg": false
  },
  "nbTranslate": {
   "displayLangs": [
    "*"
   ],
   "hotkey": "alt-t",
   "langInMainMenu": true,
   "sourceLang": "en",
   "targetLang": "fr",
   "useGoogleTranslate": true
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}