TF_Code.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "__k9NTUaW6Bx"
      },
      "source": [
        "## Overview:\n",
        "In this project, I will create a simple convolutional neural network [CNN] to classify handwriting digits using MNIST dataset using TensorFlow\n",
        "The main goal is to deploy this network to android app using TFLite\n",
        "\n",
        "\n",
        "## Requirements:\n",
        "- Python 3\n",
        "- Tensorflow v2 (I used v2.1.0)\n",
        "- Android Studio (I used the latest v3.6.3)\n",
        "\n",
        "## Find Me:\n",
        "| | |\n",
        "|-|-|\n",
        "|Mail:     |ibrahimomar357@gmail.com |\n",
        "|GitHub:   |https://github.com/IbrahimOmar91 |\n",
        "|LinkedIn: |https://www.linkedin.com/in/ibrahimomar91/ |\n",
        "|Facebook: |https://www.facebook.com/IbrahimM.Omar91 |"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Z6cAfbzSW6By"
      },
      "source": [
        "Imports"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "Z2iT5QYmW6Bz"
      },
      "outputs": [],
      "source": [
        "import numpy as np\n",
        "import tensorflow\n",
        "from tensorflow.keras.datasets import mnist\n",
        "\n",
        "from tensorflow.keras.models import Sequential\n",
        "from tensorflow.keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, InputLayer, Reshape, Dropout\n",
        "from tensorflow.keras.utils import plot_model"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "from google.colab import drive\n",
        "\n",
        "drive.mount('/content/drive')\n",
        "\n",
        "# !cp /content/drive/MyDrive/abdulla/Character/anotation_28x28/Test.csv .\n",
        "# !cp /content/drive/MyDrive/abdulla/Character/anotation_28x28/Train.csv .\n",
        "\n",
        "\n",
        "!cp /content/drive/MyDrive/abdulla/Number/anotation_28x28/Test.csv .\n",
        "!cp /content/drive/MyDrive/abdulla/Number/anotation_28x28/Train.csv .\n",
        "\n",
        "!rm -rf sample_data"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_9qaTkHiXp0t",
        "outputId": "e5dbe0a5-c957-4faf-d2e0-adec161a7154"
      },
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import pandas as pd\n",
        "\n",
        "train = pd.read_csv(\"Train.csv\")\n",
        "test = pd.read_csv(\"Test.csv\")"
      ],
      "metadata": {
        "id": "1wIDbyLAX6JW"
      },
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "dAOHFYS8W6B0"
      },
      "source": [
        "Load MNIST Dataset"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import seaborn as sns\n",
        "\n",
        "Y_train = train[\"label\"]\n",
        "\n",
        "# Drop 'label' column\n",
        "X_train = train.drop(labels = [\"label\"],axis = 1) \n",
        "\n",
        "# free some space\n",
        "del train \n",
        "\n",
        "g = sns.countplot(Y_train)\n",
        "\n",
        "Y_train.value_counts()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 539
        },
        "id": "WLudSHDRYFDg",
        "outputId": "335fde10-ed3f-4397-d298-018b5a1cc916"
      },
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.8/dist-packages/seaborn/_decorators.py:36: FutureWarning: Pass the following variable as a keyword arg: x. From version 0.12, the only valid positional argument will be `data`, and passing other arguments without an explicit keyword will result in an error or misinterpretation.\n",
            "  warnings.warn(\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0    5925\n",
              "1    5924\n",
              "2    5924\n",
              "3    5924\n",
              "4    5924\n",
              "5    5924\n",
              "6    5924\n",
              "7    5924\n",
              "8    5924\n",
              "9    5924\n",
              "Name: label, dtype: int64"
            ]
          },
          "metadata": {},
          "execution_count": 4
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Reshape image in 3 dimensions (height = 28px, width = 28px , canal = 1)\n",
        "X_train = X_train.values.reshape(-1,28,28,1)\n",
        "test = test.values.reshape(-1,28,28,1)"
      ],
      "metadata": {
        "id": "wFLhMJLgZJIw"
      },
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "# Split the train and the validation set for the fitting\n",
        "x_train, x_test, y_train, y_test = train_test_split(X_train, Y_train)"
      ],
      "metadata": {
        "id": "qZ9G31NOYGL-"
      },
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "print(x_train.shape)\n",
        "print(y_train.shape)\n",
        "\n",
        "print(x_test.shape)\n",
        "print(y_test.shape)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "jFue2rJWZPHG",
        "outputId": "4369c290-d974-4d49-943d-81c3196385a7"
      },
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(44430, 28, 28, 1)\n",
            "(44430,)\n",
            "(14811, 28, 28, 1)\n",
            "(14811,)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {
        "id": "xv-F18VMW6B0"
      },
      "outputs": [],
      "source": [
        "# (x_train, y_train), (x_test, y_test) = mnist.load_data()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4UTa5Z__W6B0"
      },
      "source": [
        "Normalize data images"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {
        "id": "0Eer0R_dW6B1"
      },
      "outputs": [],
      "source": [
        "x_train = (x_train/255.).astype(np.float32)\n",
        "x_test  = (x_test/255.).astype(np.float32)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HS7EzXXkW6B1"
      },
      "source": [
        "Create TF Model"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {
        "id": "TYmzwbHhW6B1"
      },
      "outputs": [],
      "source": [
        "model = Sequential([\n",
        "    InputLayer(input_shape=(28, 28)),\n",
        "    Reshape(target_shape=(28, 28, 1)),\n",
        "    \n",
        "    Conv2D(64, (3,3), activation='relu'),\n",
        "    MaxPooling2D((2,2)),\n",
        "    Conv2D(128, (3,3), activation='relu'),\n",
        "    MaxPooling2D((2,2)),\n",
        "    Conv2D(256, (3,3), activation='relu'),\n",
        "    MaxPooling2D((2,2)),\n",
        "    \n",
        "    Flatten(),\n",
        "    \n",
        "    Dense(512),\n",
        "    Dense(64),\n",
        "    Dense(10, activation='softmax')\n",
        "])"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 12,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "fUfOcD3tW6B2",
        "outputId": "bdaf79d1-89fe-47ad-ce73-7f13dc570978"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_1\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " reshape_1 (Reshape)         (None, 28, 28, 1)         0         \n",
            "                                                                 \n",
            " conv2d_3 (Conv2D)           (None, 26, 26, 64)        640       \n",
            "                                                                 \n",
            " max_pooling2d_3 (MaxPooling  (None, 13, 13, 64)       0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " conv2d_4 (Conv2D)           (None, 11, 11, 128)       73856     \n",
            "                                                                 \n",
            " max_pooling2d_4 (MaxPooling  (None, 5, 5, 128)        0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " conv2d_5 (Conv2D)           (None, 3, 3, 256)         295168    \n",
            "                                                                 \n",
            " max_pooling2d_5 (MaxPooling  (None, 1, 1, 256)        0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " flatten_1 (Flatten)         (None, 256)               0         \n",
            "                                                                 \n",
            " dense_3 (Dense)             (None, 512)               131584    \n",
            "                                                                 \n",
            " dense_4 (Dense)             (None, 64)                32832     \n",
            "                                                                 \n",
            " dense_5 (Dense)             (None, 10)                650       \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 534,730\n",
            "Trainable params: 534,730\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "model.summary()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "Eqts9dc_W6B2",
        "outputId": "f2d7303e-89e8-4139-8cd7-b0568a2797e6"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<IPython.core.display.Image object>"
            ]
          },
          "metadata": {},
          "execution_count": 13
        }
      ],
      "source": [
        "plot_model(model, to_file='model.png')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 14,
      "metadata": {
        "id": "AHeZfdA-W6B3"
      },
      "outputs": [],
      "source": [
        "model.compile(optimizer='adam',\n",
        "              loss='sparse_categorical_crossentropy',\n",
        "              metrics=['accuracy'])"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 15,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "D4esw1YdW6B3",
        "outputId": "a2301fe3-c55e-4a45-99bb-f479d5410cda"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/30\n",
            "87/87 [==============================] - 8s 36ms/step - loss: 0.6414 - accuracy: 0.7955 - val_loss: 0.2274 - val_accuracy: 0.9290\n",
            "Epoch 2/30\n",
            "87/87 [==============================] - 2s 24ms/step - loss: 0.1580 - accuracy: 0.9528 - val_loss: 0.1312 - val_accuracy: 0.9586\n",
            "Epoch 3/30\n",
            "87/87 [==============================] - 2s 23ms/step - loss: 0.0990 - accuracy: 0.9707 - val_loss: 0.1287 - val_accuracy: 0.9607\n",
            "Epoch 4/30\n",
            "87/87 [==============================] - 2s 22ms/step - loss: 0.0774 - accuracy: 0.9766 - val_loss: 0.0839 - val_accuracy: 0.9753\n",
            "Epoch 5/30\n",
            "87/87 [==============================] - 2s 22ms/step - loss: 0.0571 - accuracy: 0.9826 - val_loss: 0.0721 - val_accuracy: 0.9788\n",
            "Epoch 6/30\n",
            "87/87 [==============================] - 2s 24ms/step - loss: 0.0494 - accuracy: 0.9844 - val_loss: 0.0735 - val_accuracy: 0.9787\n",
            "Epoch 7/30\n",
            "87/87 [==============================] - 2s 24ms/step - loss: 0.0490 - accuracy: 0.9851 - val_loss: 0.0674 - val_accuracy: 0.9808\n",
            "Epoch 8/30\n",
            "87/87 [==============================] - 2s 25ms/step - loss: 0.0441 - accuracy: 0.9861 - val_loss: 0.0830 - val_accuracy: 0.9747\n",
            "Epoch 9/30\n",
            "87/87 [==============================] - 2s 22ms/step - loss: 0.0349 - accuracy: 0.9891 - val_loss: 0.0689 - val_accuracy: 0.9797\n",
            "Epoch 10/30\n",
            "87/87 [==============================] - 2s 22ms/step - loss: 0.0279 - accuracy: 0.9912 - val_loss: 0.0622 - val_accuracy: 0.9827\n",
            "Epoch 11/30\n",
            "87/87 [==============================] - 2s 24ms/step - loss: 0.0230 - accuracy: 0.9926 - val_loss: 0.0626 - val_accuracy: 0.9828\n",
            "Epoch 12/30\n",
            "87/87 [==============================] - 2s 24ms/step - loss: 0.0236 - accuracy: 0.9919 - val_loss: 0.0671 - val_accuracy: 0.9820\n",
            "Epoch 13/30\n",
            "87/87 [==============================] - 2s 23ms/step - loss: 0.0182 - accuracy: 0.9944 - val_loss: 0.0630 - val_accuracy: 0.9843\n",
            "Epoch 14/30\n",
            "87/87 [==============================] - 2s 25ms/step - loss: 0.0269 - accuracy: 0.9913 - val_loss: 0.0778 - val_accuracy: 0.9786\n",
            "Epoch 15/30\n",
            "87/87 [==============================] - 2s 23ms/step - loss: 0.0218 - accuracy: 0.9936 - val_loss: 0.0627 - val_accuracy: 0.9831\n",
            "Epoch 16/30\n",
            "87/87 [==============================] - 2s 22ms/step - loss: 0.0132 - accuracy: 0.9958 - val_loss: 0.0870 - val_accuracy: 0.9799\n",
            "Epoch 17/30\n",
            "87/87 [==============================] - 2s 23ms/step - loss: 0.0161 - accuracy: 0.9946 - val_loss: 0.0647 - val_accuracy: 0.9854\n",
            "Epoch 18/30\n",
            "87/87 [==============================] - 2s 23ms/step - loss: 0.0118 - accuracy: 0.9961 - val_loss: 0.0830 - val_accuracy: 0.9814\n",
            "Epoch 19/30\n",
            "87/87 [==============================] - 2s 24ms/step - loss: 0.0152 - accuracy: 0.9951 - val_loss: 0.0653 - val_accuracy: 0.9837\n",
            "Epoch 20/30\n",
            "87/87 [==============================] - 2s 24ms/step - loss: 0.0212 - accuracy: 0.9935 - val_loss: 0.0635 - val_accuracy: 0.9852\n",
            "Epoch 21/30\n",
            "87/87 [==============================] - 2s 25ms/step - loss: 0.0133 - accuracy: 0.9959 - val_loss: 0.0623 - val_accuracy: 0.9847\n",
            "Epoch 22/30\n",
            "87/87 [==============================] - 2s 23ms/step - loss: 0.0096 - accuracy: 0.9967 - val_loss: 0.0772 - val_accuracy: 0.9839\n",
            "Epoch 23/30\n",
            "87/87 [==============================] - 2s 24ms/step - loss: 0.0083 - accuracy: 0.9975 - val_loss: 0.0739 - val_accuracy: 0.9841\n",
            "Epoch 24/30\n",
            "87/87 [==============================] - 2s 24ms/step - loss: 0.0112 - accuracy: 0.9964 - val_loss: 0.0713 - val_accuracy: 0.9850\n",
            "Epoch 25/30\n",
            "87/87 [==============================] - 2s 22ms/step - loss: 0.0080 - accuracy: 0.9975 - val_loss: 0.0750 - val_accuracy: 0.9844\n",
            "Epoch 26/30\n",
            "87/87 [==============================] - 2s 24ms/step - loss: 0.0141 - accuracy: 0.9957 - val_loss: 0.0904 - val_accuracy: 0.9811\n",
            "Epoch 27/30\n",
            "87/87 [==============================] - 2s 25ms/step - loss: 0.0122 - accuracy: 0.9960 - val_loss: 0.0743 - val_accuracy: 0.9848\n",
            "Epoch 28/30\n",
            "87/87 [==============================] - 2s 23ms/step - loss: 0.0157 - accuracy: 0.9948 - val_loss: 0.0796 - val_accuracy: 0.9834\n",
            "Epoch 29/30\n",
            "87/87 [==============================] - 2s 22ms/step - loss: 0.0095 - accuracy: 0.9972 - val_loss: 0.0830 - val_accuracy: 0.9845\n",
            "Epoch 30/30\n",
            "87/87 [==============================] - 2s 24ms/step - loss: 0.0116 - accuracy: 0.9963 - val_loss: 0.0802 - val_accuracy: 0.9839\n"
          ]
        }
      ],
      "source": [
        "hist = model.fit(\n",
        "    x_train,\n",
        "    y_train,\n",
        "    epochs = 30,\n",
        "    batch_size = 512,\n",
        "    validation_data=(x_test, y_test)\n",
        ")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FOzXtT5bW6B3"
      },
      "source": [
        "### Visualise Training Process"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 16,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 297
        },
        "id": "OAtuIC4FW6B4",
        "outputId": "7a5ae4ed-a39a-47cf-b8fd-db36e2ce438b"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7fc8c6ac8a00>"
            ]
          },
          "metadata": {},
          "execution_count": 16
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "plt.xlabel('Epoch Number')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.plot(hist.history['accuracy'])\n",
        "plt.plot(hist.history['val_accuracy'])\n",
        "plt.legend(('Train','Validation'), loc='lower right')"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lZHp6wnaW6B4"
      },
      "source": [
        "### Evaluate Model"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 17,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ECnfFz4wW6B4",
        "outputId": "0e4c4743-6c6a-47a6-db51-862f5d375099"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "463/463 [==============================] - 2s 3ms/step - loss: 0.0802 - accuracy: 0.9839\n",
            "Accuracy: 98.39%\n"
          ]
        }
      ],
      "source": [
        "out = model.evaluate(x_test, y_test)\n",
        "print (\"Accuracy: {:0.2f}%\".format(out[1]*100))"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1cE31-Y9W6B5"
      },
      "source": [
        "### Convert Model to TFLite and Save"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 21,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UMIFVmOiW6B6",
        "outputId": "1969d8e2-02db-42a3-c5fb-def1798f9d75"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "WARNING:absl:Found untraced functions such as _jit_compiled_convolution_op, _jit_compiled_convolution_op, _jit_compiled_convolution_op while saving (showing 3 of 3). These functions will not be directly callable after loading.\n"
          ]
        }
      ],
      "source": [
        "converter = tensorflow.lite.TFLiteConverter.from_keras_model(model)\n",
        "tflite_model = converter.convert()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 22,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "5gakj2_wW6B6",
        "outputId": "01593125-3391-4d1a-9fce-3fb068ef458b"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "2144204"
            ]
          },
          "metadata": {},
          "execution_count": 22
        }
      ],
      "source": [
        "tflite_model_file = \"mnist_model.tflite\"\n",
        "open(tflite_model_file, 'wb').write(tflite_model)"
      ]
    }
  ],
  "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": "3.7.7"
    },
    "colab": {
      "provenance": []
    },
    "accelerator": "GPU",
    "gpuClass": "standard"
  },
  "nbformat": 4,
  "nbformat_minor": 0
}