Tutorial.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "d6d4af8f-5fce-42e5-8105-e6398516a132",
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "import hydra\n",
    "from PIL import Image\n",
    "from omegaconf import DictConfig\n",
    "import matplotlib.pyplot as plt\n",
    "from transformer_ocr.models.model import TransformerOCRCTC"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "72db2b13-b69b-4320-b41b-d5c3114d7b0a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'dataset': {'dataset': {'name': 'dmec_v6_2', 'train_annotation': 'label_ocr_v5_2_train.txt', 'valid_annotation': 'label_ocr_v5_2_test.txt', 'unchanged': {'root_dir': '/u01/vuh/datasets/OCR', 'img_height': 32, 'img_width_min': 32, 'img_width_max': 1024, 'max_readers': 8}}, 'dataloader': {'num_workers': 4, 'pin_memory': True}, 'aug': {'image_aug': True, 'masked_language_model': False}}, 'model': {'vocab': 'aAàÀảẢãÃáÁạẠăĂằẰẳẲẵẴắẮặẶâÂầẦẩẨẫẪấẤậẬbBcCdDđĐeEèÈẻẺẽẼéÉẹẸêÊềỀểỂễỄếẾệỆfFgGhHiIìÌỉỈĩĨíÍịỊjJkKlLmMnNoOòÒỏỎõÕóÓọỌôÔồỒổỔỗỖốỐộỘơƠờỜởỞỡỠớỚợỢpPqQrRsStTuUùÙủỦũŨúÚụỤưƯừỪửỬữỮứỨựỰvVwWxXyYỳỲỷỶỹỸýÝỵỴzZ0123456789!\"#$%&\\'()*+,-./:;<=>?@[\\\\]^_`{|}~ ', 'device': 'cuda:0', 'batch_size': 256, 'cnn_model': 'vgg19_bn', 'cnn_args': {'pretrained': True, 'stride_pool': [[2, 2], [2, 2], [2, 1], [2, 1], [1, 1]], 'kernel_pool': [[2, 2], [2, 2], [2, 1], [2, 1], [1, 1]], 'hidden': 144, 'dropout': 0.5}, 'transformer_type': 'conformer', 'transformer_args': {'max_seq_length': 1024, 'n_layers': 4, 'scale': True, 'd_model': 144, 'n_head': 4, 'd_feedforward': 576, 'emb_dropout': 0.0, 'pos_dropout': 0.1, 'ff_dropout': 0.1, 'conv_dropout': 0.1, 'attn_dropout': 0.1, 'activation': 'swish', 'layer_norm_eps': 1e-05, 'self_attn_type': 'abs_pos', 'half_step_residual': True, 'conv_kernel_size': 3, 'conv_expansion_factor': 2}}, 'optimizer': {'optimizer': {'type': 'adam', 'lr': 0.001, 'betas': [0.9, 0.98], 'eps': 1e-09, 'weight_decay': 0, 'lr_mul': 2.0, 'n_warm_steps': 8000}}, 'pl_params': {'pl_trainer': {'gpus': 0, 'max_epochs': 200, 'max_steps': 200000, 'num_nodes': 1, 'accelerator': 'dp', 'accumulate_grad_batches': 1, 'checkpoint_callback': True, 'log_every_n_steps': 10, 'val_every_n_steps': 500, 'val_check_interval': 0.5, 'detect_anomaly': True, 'check_val_every_n_epoch': 1, 'precision': 32, 'sync_batchnorm': False, 'benchmark': False, 'gradient_clip_val': 1.0, 'gradient_clip_algorithm': 'value'}, 'loss_func': {'blank': 1, 'reduction': 'sum', 'zero_infinity': True}, 'ctc_smoothing': 0.0, 'max_norm': 5.0, 'pretrained': 'scripts/ckpts_cf_144d/dmec', 'use_beamsearch': False, 'predict': True, 'model_callbacks': {'monitor': 'sentence accuracy', 'dirpath': 'ckpts', 'filename': 'dmec', 'save_top_k': 3, 'mode': 'max'}}, 'lm_models': {'model_path': None, 'alpha': 0.0, 'beta': 0.0, 'cutoff_top_n': 40, 'cutoff_prob': 1.0, 'beam_width': 10, 'num_processes': 8, 'blank_id': 1, 'log_probs_input': False}}"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hydra.initialize(config_path=\"scripts/hydra_cf_144d_4l\")\n",
    "config = hydra.compose(\"config.yaml\")\n",
    "config"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "c772b9aa-cfda-48c9-9fb4-ea450be888f3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "vocab_size 231\n"
     ]
    }
   ],
   "source": [
    "model = TransformerOCRCTC(config=config)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "8d07b4e8-069b-4806-8264-e3094b3386f6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7fadf6db1310>"
      ]
     },
     "execution_count": 8,
     "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": [
    "img = Image.open('test_img.png')\n",
    "plt.imshow(img)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "b577dc9c-cc6d-4243-90fa-8d092fafcaba",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(663, 24)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "img.size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "14f43e2a-47e3-4878-a2d1-739412b4033b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'This certificate is valid from November 18, 2016 to November 17, 2018'"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ocred_text = model.predict(img)\n",
    "ocred_text"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "93513832-ad0d-4c15-9d97-afd20bbd2df1",
   "metadata": {},
   "outputs": [],
   "source": [
    "# model.validation()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "02c63f6a-bfa4-4b67-8e1e-6e7404c66cdd",
   "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": "3.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}