convolutional_neural_networks.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 127,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import tensorflow as tf\n",
    "import pandas as pd\n",
    "from tensorflow.keras import models, layers\n",
    "from tensorflow.keras.utils import image_dataset_from_directory\n",
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Data storage\n",
    "DATA_FOLDER = \"./\"\n",
    "PROVIDED_TRAIN_FOLDER_NAME = \"Train\"\n",
    "PROVIDED_TEST_FOLDER_NAME = \"Test\"\n",
    "CUSTOM_TRAIN_FOLDER_NAME = \"CustomTrain\"\n",
    "CUSTOM_TEST_FOLDER_NAME = \"CustomTest\"\n",
    "DEVANAGARI_DATA_FOLDER = os.path.join(DATA_FOLDER, \"DevanagariHandwrittenDigitDataset\")\n",
    "\n",
    "RANDOM_SEED = 20\n",
    "\n",
    "# Data parameters\n",
    "BATCH_SIZE = 32\n",
    "IMG_HEIGHT = 32\n",
    "IMG_WIDTH = 32\n",
    "\n",
    "# Training parameters\n",
    "epochs = 15\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "metadata": {},
   "outputs": [],
   "source": [
    "nums_to_labels = {0: 'digit_0', 1: 'digit_1', 2: 'digit_2', 3: 'digit_3', 4: 'digit_4', 5: 'digit_5', 6: 'digit_6', 7: 'digit_7', 8: 'digit_8', 9: 'digit_9'}\n",
    "label_names = ['digit_0', 'digit_1', 'digit_2', 'digit_3', 'digit_4', 'digit_5', 'digit_6', 'digit_7', 'digit_8', 'digit_9']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 17000 files belonging to 10 classes.\n"
     ]
    }
   ],
   "source": [
    "train_dataset = image_dataset_from_directory(\n",
    "    os.path.join(DEVANAGARI_DATA_FOLDER, PROVIDED_TRAIN_FOLDER_NAME),\n",
    "    labels=\"inferred\",\n",
    "    label_mode=\"int\",\n",
    "    class_names=label_names,\n",
    "    color_mode=\"grayscale\",\n",
    "    batch_size=BATCH_SIZE,\n",
    "    image_size=(IMG_HEIGHT, IMG_WIDTH),\n",
    "    shuffle=True,\n",
    "    seed=RANDOM_SEED,\n",
    "    validation_split=None, # None, so that we get all the data.\n",
    "    subset=None,\n",
    "    interpolation=\"bilinear\",\n",
    "    follow_links=False,\n",
    "    crop_to_aspect_ratio=False,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 80 files belonging to 10 classes.\n"
     ]
    }
   ],
   "source": [
    "custom_train_dataset = image_dataset_from_directory(\n",
    "    os.path.join(DEVANAGARI_DATA_FOLDER, CUSTOM_TRAIN_FOLDER_NAME),\n",
    "    labels=\"inferred\",\n",
    "    label_mode=\"int\",\n",
    "    class_names=label_names,\n",
    "    color_mode=\"grayscale\",\n",
    "    batch_size=BATCH_SIZE,\n",
    "    image_size=(IMG_HEIGHT, IMG_WIDTH),\n",
    "    shuffle=True,\n",
    "    seed=RANDOM_SEED,\n",
    "    validation_split=None, # None, so that we get all the data.\n",
    "    subset=None,\n",
    "    interpolation=\"bilinear\",\n",
    "    follow_links=False,\n",
    "    crop_to_aspect_ratio=False,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 3000 files belonging to 10 classes.\n"
     ]
    }
   ],
   "source": [
    "# Provided test dataset\n",
    "test_dataset = image_dataset_from_directory(\n",
    "    os.path.join(DEVANAGARI_DATA_FOLDER, PROVIDED_TEST_FOLDER_NAME),\n",
    "    labels=\"inferred\",\n",
    "    label_mode=\"int\",\n",
    "    class_names=label_names,\n",
    "    color_mode=\"grayscale\",\n",
    "    batch_size=BATCH_SIZE,\n",
    "    image_size=(IMG_HEIGHT, IMG_WIDTH),\n",
    "    shuffle=True,\n",
    "    seed=RANDOM_SEED,\n",
    "    validation_split=None,  # None, so that we get all the data.\n",
    "    subset=None,\n",
    "    interpolation=\"bilinear\",\n",
    "    follow_links=False,\n",
    "    crop_to_aspect_ratio=False,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 100 files belonging to 10 classes.\n"
     ]
    }
   ],
   "source": [
    "# Provided test dataset\n",
    "\n",
    "custom_test_dataset = image_dataset_from_directory(\n",
    "    os.path.join(DEVANAGARI_DATA_FOLDER, CUSTOM_TEST_FOLDER_NAME),\n",
    "    labels=\"inferred\",\n",
    "    label_mode=\"int\",\n",
    "    class_names=label_names,\n",
    "    color_mode=\"grayscale\",\n",
    "    batch_size=BATCH_SIZE,\n",
    "    image_size=(IMG_HEIGHT, IMG_WIDTH),\n",
    "    shuffle=True,\n",
    "    seed=RANDOM_SEED,\n",
    "    validation_split=None,  # None, so that we get all the data.\n",
    "    subset=None,\n",
    "    interpolation=\"bilinear\",\n",
    "    follow_links=False,\n",
    "    crop_to_aspect_ratio=False,\n",
    ")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(32, 32, 1)\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1200x1200 with 9 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Take a look at the input data\n",
    "rows = 3\n",
    "columns = 3\n",
    "fig = plt.figure(figsize=(12, 12))\n",
    "i = 1\n",
    "for images, labels in train_dataset.take(1):\n",
    "    print(images[0].shape)\n",
    "    for image, label in zip(images[:9], labels[:9]):\n",
    "        fig.add_subplot(rows, columns, i)\n",
    "        plt.imshow(image, cmap='gray', vmin=0, vmax=255)\n",
    "        plt.title(nums_to_labels[int(label)])\n",
    "        i += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/15\n",
      "538/538 [==============================] - 5s 9ms/step - loss: 0.5214 - accuracy: 0.8945 - val_loss: 0.2118 - val_accuracy: 0.9350\n",
      "Epoch 2/15\n",
      "538/538 [==============================] - 4s 8ms/step - loss: 0.0758 - accuracy: 0.9767 - val_loss: 0.0729 - val_accuracy: 0.9793\n",
      "Epoch 3/15\n",
      "538/538 [==============================] - 4s 8ms/step - loss: 0.0456 - accuracy: 0.9862 - val_loss: 0.0641 - val_accuracy: 0.9847\n",
      "Epoch 4/15\n",
      "538/538 [==============================] - 4s 8ms/step - loss: 0.0336 - accuracy: 0.9906 - val_loss: 0.0932 - val_accuracy: 0.9797\n",
      "Epoch 5/15\n",
      "538/538 [==============================] - 4s 8ms/step - loss: 0.0291 - accuracy: 0.9909 - val_loss: 0.0703 - val_accuracy: 0.9833\n",
      "Epoch 6/15\n",
      "538/538 [==============================] - 4s 8ms/step - loss: 0.0315 - accuracy: 0.9898 - val_loss: 0.0501 - val_accuracy: 0.9870\n",
      "Epoch 7/15\n",
      "538/538 [==============================] - 4s 8ms/step - loss: 0.0190 - accuracy: 0.9941 - val_loss: 0.0570 - val_accuracy: 0.9887\n",
      "Epoch 8/15\n",
      "538/538 [==============================] - 4s 8ms/step - loss: 0.0090 - accuracy: 0.9973 - val_loss: 0.0763 - val_accuracy: 0.9860\n",
      "Epoch 9/15\n",
      "538/538 [==============================] - 4s 8ms/step - loss: 0.0342 - accuracy: 0.9902 - val_loss: 0.0514 - val_accuracy: 0.9920\n",
      "Epoch 10/15\n",
      "538/538 [==============================] - 5s 8ms/step - loss: 0.0102 - accuracy: 0.9967 - val_loss: 0.0832 - val_accuracy: 0.9843\n",
      "Epoch 11/15\n",
      "538/538 [==============================] - 5s 9ms/step - loss: 0.0315 - accuracy: 0.9920 - val_loss: 0.0388 - val_accuracy: 0.9923\n",
      "Epoch 12/15\n",
      "538/538 [==============================] - 4s 8ms/step - loss: 0.0128 - accuracy: 0.9966 - val_loss: 0.0360 - val_accuracy: 0.9940\n",
      "Epoch 13/15\n",
      "538/538 [==============================] - 4s 8ms/step - loss: 0.0160 - accuracy: 0.9957 - val_loss: 0.0604 - val_accuracy: 0.9907\n",
      "Epoch 14/15\n",
      "538/538 [==============================] - 4s 8ms/step - loss: 0.0093 - accuracy: 0.9977 - val_loss: 0.0559 - val_accuracy: 0.9893\n",
      "Epoch 15/15\n",
      "538/538 [==============================] - 5s 9ms/step - loss: 0.0191 - accuracy: 0.9947 - val_loss: 0.0583 - val_accuracy: 0.9900\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x288801e40>"
      ]
     },
     "execution_count": 141,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = models.Sequential(\n",
    "    [\n",
    "       \n",
    "        layers.Conv2D(BATCH_SIZE, 3, activation=\"relu\", input_shape=(IMG_WIDTH, IMG_HEIGHT, 1)),\n",
    "        layers.MaxPooling2D((2,2)),\n",
    "        layers.Conv2D(BATCH_SIZE*2, 3, activation=\"relu\"),\n",
    "        layers.MaxPooling2D((2,2)),\n",
    "        layers.Conv2D(BATCH_SIZE*2, 3, activation=\"relu\"),\n",
    "        layers.Flatten(),\n",
    "        layers.Dense(BATCH_SIZE*2, activation=\"relu\"),\n",
    "        layers.Dense(10, activation=\"softmax\")\n",
    "    ]\n",
    ")\n",
    "train_dataset = train_dataset.concatenate(custom_train_dataset)\n",
    "model.compile(optimizer=\"adam\", loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True), metrics=[\"accuracy\"])\n",
    "model.fit(train_dataset, epochs=epochs, validation_data=test_dataset)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 142,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "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"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: hindi_digit_recognizer.model/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: hindi_digit_recognizer.model/assets\n"
     ]
    }
   ],
   "source": [
    "model.save(\"hindi_digit_recognizer.model\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 143,
   "metadata": {},
   "outputs": [],
   "source": [
    "del model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 144,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = models.load_model(\"./hindi_digit_recognizer.model\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 26/538 [>.............................] - ETA: 2s - loss: 0.0081 - accuracy: 0.9964   "
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Library/Frameworks/Python.framework/Versions/3.10/lib/python3.10/site-packages/keras/backend.py:5582: UserWarning: \"`sparse_categorical_crossentropy` received `from_logits=True`, but the `output` argument was produced by a Softmax activation and thus does not represent logits. Was this intended?\n",
      "  output, from_logits = _get_logits(\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "538/538 [==============================] - 2s 4ms/step - loss: 0.0127 - accuracy: 0.9961\n",
      "94/94 [==============================] - 0s 4ms/step - loss: 0.0583 - accuracy: 0.9900\n",
      "Loss: 0.01272672601044178\n",
      "Accuracy: 0.996095597743988\n",
      "Test loss: 0.058316465467214584\n",
      "Test accuracy: 0.9900000095367432\n"
     ]
    }
   ],
   "source": [
    "loss, accuracy = model.evaluate(train_dataset)\n",
    "test_loss, test_accuracy = model.evaluate(test_dataset)\n",
    "print(f\"Loss: {loss}\\nAccuracy: {accuracy}\")\n",
    "print(f\"Test loss: {test_loss}\\nTest accuracy: {test_accuracy}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4/4 [==============================] - 0s 3ms/step - loss: 2.3050 - accuracy: 0.8400\n",
      "Custom test loss: 2.3050475120544434\n",
      "Custom test accuracy: 0.8399999737739563\n"
     ]
    }
   ],
   "source": [
    "custom_test_loss, custom_test_accuracy = model.evaluate(custom_test_dataset)\n",
    "print(f\"Custom test loss: {custom_test_loss}\\nCustom test accuracy: {custom_test_accuracy}\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.10.8 64-bit",
   "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": "3.10.8"
  },
  "vscode": {
   "interpreter": {
    "hash": "7e1998ff7f8aa20ada591c520b972326324e5ea05489af9e422744c7c09f6dad"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}