demo/crnn_iiit5k.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Intro"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<img src=\"images/iiit5k.png\" width=\"600\">\n",
    "<p>This notebook demonstrates training and testing a CRNN model on the <a href=\"https://cvit.iiit.ac.in/projects/SceneTextUnderstanding/IIIT5K.html\">IIIT 5K-word dataset</a>. As the picture above shows, this dataset contains images with some text and the true text for each of these images.</p>\n",
    "<p>The CRNN architecture for the image-based sequence recognition was introduced in <a href=\"https://arxiv.org/abs/1507.05717\">(Shi et al., 2015)</a>. In this paper authors train the model on a large dataset of synthetically generated images with text and then test it on various other datasets including the IIIT5K. However, in this notebook we will both train and test the model solely on the IIIT5K for the sake of simplicity.</p>\n",
    "<p>We will first explore a simpler way of creating a CRNN as a Sequential model, that doesn't take in account a width of an image (we can make this work with the IIT5K by padding all the images to the same width) and then we will also wrap it in another model, that accepts image width as one of its inputs.</p>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Import dependencies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from pathlib import Path\n",
    "from mat4py import loadmat\n",
    "from PIL import Image\n",
    "import os\n",
    "\n",
    "import sys\n",
    "sys.path.insert(0, \"..\")\n",
    "import crnn\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "AUTOTUNE = tf.data.experimental.AUTOTUNE"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Download and preprocess the dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "First, let's download and extract the dataset:<br/>\n",
    "(**Warning: 106 MB download!**)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "dataset_path = tf.keras.utils.get_file(\n",
    "    \"IIIT5K-Word_V3.0.tar.gz\",\n",
    "    \"http://cvit.iiit.ac.in/projects/SceneTextUnderstanding/IIIT5K-Word_V3.0.tar.gz\",\n",
    "    extract=True)\n",
    "dataset_path = Path(dataset_path).parent / \"IIIT5K\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The dataset provides true text for each image in a form of dictionaries, so that image's filename can be used as a key to retreive the true text for this image. The dictionaries come in a form of MATLAB files, so we wil use the `mat4py` package to load them (loading might take some time):"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_filename_to_text_dict(dataset_path, name):\n",
    "    filename_to_text = loadmat((dataset_path / f\"{name}.mat\").as_posix())[name]\n",
    "    return {Path(path).name: text\n",
    "            for path, text in zip(filename_to_text[\"ImgName\"],\n",
    "                                  filename_to_text[\"GroundTruth\"])}\n",
    "\n",
    "filename_to_text_train = load_filename_to_text_dict(dataset_path, \"traindata\")\n",
    "filename_to_text_test = load_filename_to_text_dict(dataset_path, \"testdata\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Since we have true text for all images, we can now create an array of all unique characters:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['0', '1', '2', '3', '4', '5', '6', '7', '8', '9', 'A', 'B', 'C',\n",
       "       'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P',\n",
       "       'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z'], dtype='<U1')"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unique_chars = np.asarray(\n",
    "    sorted({c for text in filename_to_text_train.values() for c in text}))\n",
    "unique_chars"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To make the dictionaries usable from TensorFlow graphs, we need to convert them to `StaticHashTable`'s:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def dict_to_tf(dictionary, default_value=\"\"):\n",
    "    keys = list(dictionary.keys())\n",
    "    values = [dictionary[key] for key in keys]\n",
    "    return tf.lookup.StaticHashTable(\n",
    "        tf.lookup.KeyValueTensorInitializer(keys, values), default_value)\n",
    "\n",
    "filename_to_text_train = dict_to_tf(filename_to_text_train)\n",
    "filename_to_text_test = dict_to_tf(filename_to_text_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's plot some images from the train dataset with their texts from the corresponding dictionary to check that everything works as expected:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1224x72 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot_images(images, text, rows, columns, cmap=None, norm=None):\n",
    "    for i in range(len(images)):\n",
    "        image = images[i]\n",
    "        plt.subplot(rows, columns, i + 1)\n",
    "        plt.title(text[i], {'fontsize': 18})\n",
    "        plt.imshow(image, cmap=cmap, norm=norm)\n",
    "\n",
    "image_paths = list((dataset_path / \"train\").glob(\"*.png\"))[:4]\n",
    "images = [np.asarray(Image.open(path)) for path in image_paths]\n",
    "text = [filename_to_text_train.lookup(tf.constant(path.name)).numpy().decode(\"UTF-8\")\n",
    "        for path in image_paths]\n",
    "\n",
    "plt.figure(figsize=(17, 1))\n",
    "plot_images(images, text, 1, 4)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Knowing the unique characters we can create `StaticHashTable`'s to encode characters with numeric labels and to decode them back:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels = np.arange(unique_chars.size, dtype=np.int32)\n",
    "char_to_label = tf.lookup.StaticHashTable(\n",
    "    tf.lookup.KeyValueTensorInitializer(unique_chars, labels), -1)\n",
    "label_to_char = tf.lookup.StaticHashTable(\n",
    "    tf.lookup.KeyValueTensorInitializer(labels, unique_chars), \"\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We also will need a function to split a tensor into separate characters before encoding:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "def split_chars(tensor, encoding=\"UTF-8\"):\n",
    "    return tf.strings.unicode_split(tensor, encoding)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's encode and decode back an example word:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "A source word: TENSORFLOW\n",
      "An encoded word: [29 14 23 28 24 27 15 21 24 32]\n",
      "A decoded word: TENSORFLOW\n"
     ]
    }
   ],
   "source": [
    "source_word = split_chars(\"TENSORFLOW\", \"UTF-8\")\n",
    "print(\"A source word:\", b\"\".join(source_word.numpy()).decode(\"utf-8\"))\n",
    "\n",
    "encoded_word = char_to_label.lookup(source_word)\n",
    "print(\"An encoded word:\", encoded_word.numpy())\n",
    "\n",
    "decoded_word = label_to_char.lookup(encoded_word)\n",
    "print(\"A decoded word:\",  b\"\".join(decoded_word.numpy()).decode(\"utf-8\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we can build train and test datasets using the `tf.data` API. This is a pretty standard images loading routine, but there are still a couple of things to note:\n",
    "- All images are scaled to have the same height (32 pixels)\n",
    "- The resulting datasets contain image's width in addition to the image itself and its true text (we will use the width for the second model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "IMAGE_CHANNELS = 1 # Images will be converted to grayscale\n",
    "IMAGE_HEIGHT = 32\n",
    "\n",
    "rescaling = tf.keras.layers.experimental.preprocessing.Rescaling(\n",
    "    scale=1./127.5, offset=-1)\n",
    "\n",
    "def build_dataset(path, filename_to_text, char_to_label):\n",
    "    files_ds = tf.data.Dataset.list_files((path / \"*.png\").as_posix(), shuffle=False)\n",
    "    return files_ds.map(\n",
    "        lambda image_path: process_path(image_path, filename_to_text, char_to_label),\n",
    "        num_parallel_calls=AUTOTUNE)\n",
    "\n",
    "def process_path(path, filename_to_text, char_to_label):\n",
    "    image = load_image(path)\n",
    "    image_width = tf.shape(image)[1]\n",
    "    labels = get_labels(path, filename_to_text, char_to_label)\n",
    "    return image, image_width, labels\n",
    "\n",
    "def load_image(path):\n",
    "    image_data = tf.io.read_file(path)\n",
    "    image = tf.image.decode_png(image_data, channels=IMAGE_CHANNELS)\n",
    "    image = rescaling(image)\n",
    "    \n",
    "    # Scaling image to the specified height\n",
    "    source_shape = tf.shape(image)\n",
    "    scale_factor = IMAGE_HEIGHT / tf.cast(source_shape[0], tf.float32)\n",
    "    new_width = tf.cast(source_shape[1], tf.float32) * scale_factor\n",
    "    image = tf.image.resize(image, [IMAGE_HEIGHT, new_width])\n",
    "    \n",
    "    return image\n",
    "\n",
    "def get_labels(path, filename_to_text, char_to_label):\n",
    "    image_name = tf.strings.split(path, os.path.sep)[-1]\n",
    "    image_text = filename_to_text.lookup(image_name)\n",
    "    labels = char_to_label.lookup(split_chars(image_text))\n",
    "    return labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train dataset size: 2000\n",
      "Test dataset size: 3000\n"
     ]
    }
   ],
   "source": [
    "source_train_ds = build_dataset(\n",
    "    dataset_path / \"train\", filename_to_text_train, char_to_label)\n",
    "source_test_ds = build_dataset(\n",
    "    dataset_path / \"test\", filename_to_text_test, char_to_label)\n",
    "\n",
    "print(\"Train dataset size:\", tf.data.experimental.cardinality(source_train_ds).numpy())\n",
    "print(\"Test dataset size:\", tf.data.experimental.cardinality(source_test_ds).numpy())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<p>As we can see, the default train/test split of the dataset puts the majority of the images into the test set, while normally we would like to do it vice versa. Moreover, distributions of train and test sets seem to be different in some way, at least when using them both to test a model that was trained on another dataset, it performs significantly better on the train set even though it wasn't actually trained on it.</p>\n",
    "<p>This factors limit a performance of a model, that is trained with the default train/test split. To fix this, both sets are concatenated together and shuffled to fix the difference of their distributions, and then they are splitted into train and test sets with a more traditional 80%/20% ratio.</p>\n",
    "<p><b>Note:</b> If you are using the IIIT5K dataset to benchmark your model, you should stick to the default split in order to keep your result consistent with other models that were tested using the same dataset. However, in this notebook we are using the dataset just to demonstrate the training process, so there is no problem in using a custom split.</p>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "New train dataset size: 4000\n",
      "New test dataset size: 1000\n"
     ]
    }
   ],
   "source": [
    "reshuffled_source_ds = source_train_ds.concatenate(source_test_ds).shuffle(\n",
    "    5000, seed=42, reshuffle_each_iteration=False)\n",
    "source_train_ds = reshuffled_source_ds.take(4000)\n",
    "source_test_ds = reshuffled_source_ds.skip(4000)\n",
    "\n",
    "print(\"New train dataset size:\", tf.data.experimental.cardinality(source_train_ds).numpy())\n",
    "print(\"New test dataset size:\", tf.data.experimental.cardinality(source_test_ds).numpy())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<p>Finally, we need to batch our datasets and configure them for performance. This is another common procedure. The only special thing here is that all images have different width (height is the same since we scaled them before) and their true text consists of different number of characters. To be able to batch tensors of different shape together, we need to use the `padded_batch()` function.</p>\n",
    "<p>We could just batch everything to the maximum value in the batch, but having images tensors of different width in each batch will cause a model's graph to be retraced every time it gets a batch with a new width, drastically slowing down the model's training. To avoid this, we calculate the maximum image width in the whole dataset and pad all images to that width.</p>\n",
    "<p>For the true text, however, this is not required, since loss and accuracy are not retraced for each new length of the true labels sequence, so we can just allow the `padded_batch()` to use the maximum length in each batch. And, of course, since image width value is just a scalar for all images, there is no need to pad it at all.</p>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "SHUFFLE_BUFFER_SIZE = 1000\n",
    "BATCH_SIZE = 32\n",
    "IMAGE_PADDING_CONST = 0.0\n",
    "LABELS_PADDING_CONST = -1\n",
    "MAX_IMAGE_WIDTH = max(width.numpy() for _, width, _ in source_train_ds.concatenate(source_test_ds))\n",
    "\n",
    "def configure_for_performance(dataset, shuffle=False, seed=None, ):\n",
    "    dataset = dataset.cache()\n",
    "    if shuffle:\n",
    "      dataset = dataset.shuffle(buffer_size=SHUFFLE_BUFFER_SIZE, seed=seed)\n",
    "    dataset = dataset.padded_batch(\n",
    "        BATCH_SIZE, padding_values=(IMAGE_PADDING_CONST, None, LABELS_PADDING_CONST),\n",
    "        padded_shapes=([IMAGE_HEIGHT, MAX_IMAGE_WIDTH, IMAGE_CHANNELS], [], [None],)\n",
    "        )\n",
    "    dataset = dataset.prefetch(buffer_size=AUTOTUNE)\n",
    "    return dataset\n",
    "\n",
    "source_train_ds = configure_for_performance(source_train_ds, shuffle=True, seed=42)\n",
    "source_test_ds = configure_for_performance(source_test_ds) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's plot some images with their true text from our final training dataset to make sure that everything is alright:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x648 with 5 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "images, _, labels = next(iter(source_train_ds))\n",
    "images, labels = images[:5], labels[:5]\n",
    "images = [(np.squeeze(image) + 1) / 2 for image in images.numpy()]\n",
    "text = [\"\".join(c.decode(\"UTF-8\") for c in chars)\n",
    "        for chars in label_to_char.lookup(labels).numpy()]\n",
    "    \n",
    "plt.figure(figsize=(15, 9))\n",
    "plot_images(images, text, rows=5, columns=1, \n",
    "            cmap=\"binary\", norm=matplotlib.colors.Normalize(-1, 1))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Sequential model for fixed-length input"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<p>We will start with building a CRNN as a standard Sequential model. This approach is easy and intuitive, however such model won't take in account the input image's width when calculating loss and accuracy, so it is most suitable for problems with constant-width images like, for example, CAPTCHA solving or license plate recognition.</p>\n",
    "<p>However, we can still use it for the IIIT5K dataset since we have padded all images to the same width. It is worth mentioning, that I haven't actually noticed any significant difference in terms of training speed or accuracy between this model and the one, that takes in account the image's width (we will also train it later). This is probably because the model easily learns to just predict the blank character for all empty areas of images that were added during the padding process.</p>\n",
    "<p>Thus, I would suggest to try this simpler model even if your dataset contains images with different width, especially if the width doesn't vary to much.</p>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "First, we need to exlude image width value from the source datasets to use them for this model:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_ds = source_train_ds.map(lambda image, width, text: (image, text))\n",
    "test_ds = source_test_ds.map(lambda image, width, text: (image, text))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we can build the model. Its architecture is similar to the original paper referenced in the [Intro](#Intro), with some dropout and L2 regularization added to reduce overfitting:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "CONV_DROPOUT = 0.35\n",
    "LSTM_L2 = 0.01\n",
    "DENSE_DROPOUT = 0.5\n",
    "DENSE_L2 = 0.01\n",
    "\n",
    "def build_model():\n",
    "    return tf.keras.Sequential([\n",
    "        tf.keras.layers.Conv2D(64, (3, 3), activation=\"relu\", padding=\"same\"),\n",
    "        tf.keras.layers.MaxPool2D((2, 2)),\n",
    "        tf.keras.layers.SpatialDropout2D(CONV_DROPOUT),\n",
    "\n",
    "        tf.keras.layers.Conv2D(128, (3, 3), activation=\"relu\", padding=\"same\"),\n",
    "        tf.keras.layers.MaxPool2D((2, 2)),\n",
    "        tf.keras.layers.SpatialDropout2D(CONV_DROPOUT),\n",
    "\n",
    "        tf.keras.layers.Conv2D(256, (3, 3), activation=\"relu\", padding=\"same\"),\n",
    "        tf.keras.layers.Conv2D(256, (3, 3), activation=\"relu\", padding=\"same\"),\n",
    "        tf.keras.layers.MaxPool2D((2, 1)),\n",
    "        tf.keras.layers.SpatialDropout2D(CONV_DROPOUT),\n",
    "\n",
    "        tf.keras.layers.Conv2D(512, (3, 3), activation=\"relu\", padding=\"same\"),\n",
    "        tf.keras.layers.BatchNormalization(),\n",
    "        tf.keras.layers.Conv2D(512, (3, 3), activation=\"relu\", padding=\"same\"),\n",
    "        tf.keras.layers.BatchNormalization(),\n",
    "        tf.keras.layers.MaxPool2D((2, 1)),\n",
    "        tf.keras.layers.SpatialDropout2D(CONV_DROPOUT),\n",
    "\n",
    "        tf.keras.layers.Conv2D(512, (2, 2), activation=\"relu\"),\n",
    "        tf.keras.layers.SpatialDropout2D(CONV_DROPOUT),\n",
    "\n",
    "        crnn.layers.MapToSequence(),\n",
    "\n",
    "        tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(256, return_sequences=True,\n",
    "                                      kernel_regularizer=tf.keras.regularizers.l2(LSTM_L2))),\n",
    "        tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(256, return_sequences=True,\n",
    "                                      kernel_regularizer=tf.keras.regularizers.l2(LSTM_L2))),\n",
    "\n",
    "        tf.keras.layers.Dropout(DENSE_DROPOUT),\n",
    "\n",
    "        tf.keras.layers.Dense(unique_chars.size + 1, activation=\"softmax\",\n",
    "                             kernel_regularizer=tf.keras.regularizers.l2(DENSE_L2))\n",
    "    ])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "model = build_model()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "When compile the model we use `CTCLoss` and `CTCAccuracy`, which calculates how often a predicted text fully matches the corresponding true text:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "model.compile(\n",
    "    optimizer=tf.keras.optimizers.Adam(learning_rate=0.0001), \n",
    "    loss=crnn.losses.CTCLoss(y_true_padding_const=LABELS_PADDING_CONST),\n",
    "    metrics=[crnn.metrics.CTCAccuracy(y_true_padding_const=LABELS_PADDING_CONST)]\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's also define some callbacks to automate the training:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "callbacks = [\n",
    "    tf.keras.callbacks.ReduceLROnPlateau(\n",
    "        factor=0.5, patience=8, min_delta=0, verbose=1),\n",
    "    tf.keras.callbacks.EarlyStopping(\n",
    "        patience=20, verbose=1, restore_best_weights=True)\n",
    "]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we are ready to train the model. Please note that training process takes a substantial amount of time even using a GPU, so you might want to skip it and just load the saved weights by uncommenting the code in the cell below."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "model.fit(train_ds, validation_data=test_ds, epochs=300, callbacks=callbacks)\n",
    "model.save_weights(Path(\"saved_weights\") / \"crnn_iiit5k.h5\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Uncomment to load pretrained \n",
    "#model.build((None, IMAGE_HEIGHT, None, IMAGE_CHANNELS))\n",
    "#model.load_weights(Path(\"saved_weights\") / \"crnn_iiit5k.h5\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's evaluate the model:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "125/125 [==============================] - 5s 43ms/step - loss: 0.4891 - accuracy: 1.0000\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.4890742599964142, 1.0]"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(train_ds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "32/32 [==============================] - 1s 42ms/step - loss: 2.4263 - accuracy: 0.7870\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[2.4263124465942383, 0.7870000004768372]"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(test_ds)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The model correctly recognizes text on 78.7% of the test images. The original paper reports 78.2% accuracy, but the authors used the IIIT5K dataset only for evaluation and trained their model on another dataset, so there is no direct comparison. The model overfits, however it is not as bad as it would be without dropout and L2 regularization. Variance can possibly be reduced even more by further increasing dropout and/or regularization rates, using the weight decay instead of the L2 regularization (since for the Adam optimizer they are not equivalent), etc."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Currently the model outputs CTC labels, but we can append a CTCDecode layer to it before using it for inference:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = tf.keras.Sequential([\n",
    "    model,\n",
    "    crnn.layers.CTCDecode(label_to_char)\n",
    "])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's now try out the model with the decoding layer:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "image = Image.open(\"images/tensorflow.png\").convert(mode=\"L\")\n",
    "image = image.resize((round(image.size[0] * IMAGE_HEIGHT / image.size[1]), IMAGE_HEIGHT),\n",
    "                     resample=Image.BILINEAR)\n",
    "image = np.asarray(image)[np.newaxis, :, :, np.newaxis]\n",
    "image = rescaling(image)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = model(image)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "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"
    }
   ],
   "source": [
    "text = b\"\".join(y_pred[0].numpy()).decode(\"UTF-8\")\n",
    "plot_images(image[:, :, :, 0], [text], rows=1, columns=1, \n",
    "            cmap=\"binary\", norm=matplotlib.colors.Normalize(-1, 1))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Well, the model got everything right with this image!"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# The Functional API for a variable-width images"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this section we will build a model using the Functional API, that takes image's width as an input and passes it to loss and accuracy."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For this model we need to place tensors from the source datasets into a nested tuple, so that they all will be passed as inputs to the model:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_ds = source_train_ds.map(lambda image, width, text: ((image, width, text),))\n",
    "test_ds = source_test_ds.map(lambda image, width, text: ((image, width, text),))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The same sequential model will be used as a \"core\" model:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "core_model = build_model()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will wrap the core model in another model built with the Functional API. It will take an image, its width and its true text as inputs, \"convert\" the image to predictions using the core model and then pass all this three tensors to loss and accuracy layers. Note that loss and accuracy layers are \"transparent\" (they return the same predictions that were passed to them), so we can just stack them after the core model. After training this new model, you can use just the core model for inference."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "image_input = tf.keras.layers.Input((IMAGE_HEIGHT, None, IMAGE_CHANNELS), name=\"image_input\")\n",
    "image_width_input = tf.keras.layers.Input(tuple(), name=\"image_width_input\")\n",
    "labels_input = tf.keras.layers.Input((None,), name=\"labels_input\")\n",
    "\n",
    "y_true = labels_input\n",
    "y_pred = core_model(image_input)\n",
    "\n",
    "y_pred = crnn.layers.CTCLossLayer(y_true_padding_const=LABELS_PADDING_CONST)(\n",
    "    y_true, y_pred, image_input, image_width_input)\n",
    "y_pred = crnn.layers.CTCAccuracyLayer(y_true_padding_const=LABELS_PADDING_CONST)(\n",
    "    y_true, y_pred, image_input, image_width_input)\n",
    "\n",
    "model = tf.keras.Model(inputs=[image_input, image_width_input, labels_input],\n",
    "                       outputs=y_pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Since we've already added loss and accuracy layers, only an optimizer need to be specified when compiling the model: "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=0.0001))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we are all set up to train the model:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "model.fit(train_ds, validation_data=test_ds, epochs=300, callbacks=callbacks)\n",
    "model.save_weights(Path(\"saved_weights\") / \"crnn_iiit5k_var_width.h5\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Uncomment to load pretrained \n",
    "#model.build((None, IMAGE_HEIGHT, None, IMAGE_CHANNELS))\n",
    "#model.load_weights(Path(\"saved_weights\") / \"crnn_iiit5k_var_width.h5\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's evaluate the model:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "125/125 [==============================] - 5s 43ms/step - loss: 0.4332 - accuracy: 1.0000\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.4332295060157776, 1.0]"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(train_ds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "32/32 [==============================] - 1s 42ms/step - loss: 2.2707 - accuracy: 0.7980\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[2.2706708908081055, 0.7979999780654907]"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(test_ds)"
   ]
  }
 ],
 "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}