imageCollection.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import cv2 \n",
    "import uuid\n",
    "import os\n",
    "import time\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt \n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Reading data from .csv files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = pd.read_csv('Tensorflow/workspace/images/train/sign_mnist_train.csv')\n",
    "test = pd.read_csv('Tensorflow/workspace/images/test/sign_mnist_test.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "train.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Adding label and calculating number of datasets for each label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels = train['label'].values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "label_array=np.array(labels)\n",
    "np.unique(label_array)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.figure(figsize=(20,10))\n",
    "sns.countplot(x=labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Above we can see that each label 0-24 has around 900-1300 images"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Refining the data to such that only pixels are left i.e removing labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "train.drop('label',axis = 1,inplace = True)\n",
    "train.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Extracting Images data from each row in our .csv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "images= train.values \n",
    "images= np.array([np.reshape(i,(28,28)) for i in images])\n",
    "images= np.array([i.flatten() for i in images])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Hot one encoding the labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!pip3 install scikit-learn\n",
    "from sklearn.preprocessing import LabelBinarizer\n",
    "\n",
    "label_binar = LabelBinarizer()\n",
    "labels = label_binar.fit_transform(labels)\n",
    "labels"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "View images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.imshow(images[33].reshape(28,28))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.imshow(images[2].reshape(28,28))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Spliting our dataset into x_train, x_test, y_train, y_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "x_train, x_test, y_train, y_test = train_test_split(images, labels, test_size= 0.3, random_state=101)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Scale our images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train = x_train/255\n",
    "x_test = x_test/255"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Reshape images to the size required by Tensorflow and Keras"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train = x_train.reshape(x_train.shape[0],28,28,1)\n",
    "x_test = x_test.reshape(x_test.shape[0],28,28,1)\n",
    "\n",
    "plt.imshow(x_train[0].reshape(28,28))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Building our own dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "class_names = ['A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'K', 'L', 'M', 'N', 'O', 'P','Q', 'R', 'S','T','U','V','W','X','Y']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Function to setup the directories\n",
    "\n",
    "IMAGES_PATH = os.path.join('Tensorflow', 'workspace', 'images')\n",
    "IMAGES_PATH\n",
    "os.name"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "if not os.path.exists(IMAGES_PATH):\n",
    "    if os.name == 'posix':\n",
    "        !mkdir -p {IMAGES_PATH}\n",
    "    if os.name == 'nt':\n",
    "        !mkdir {IMAGES_PATH}\n",
    "\n",
    "for name in class_names:\n",
    "    path = os.path.join(IMAGES_PATH, name)\n",
    "    if not os.path.exists(path):\n",
    "        !mkdir {path}\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "for name in class_names: \n",
    "    cap = cv2.VideoCapture(0) #Connects to our webcam; for could be 2 instead of 0\n",
    "    if cap.isOpened():\n",
    "        print('Collecting images for {}'.format(name))\n",
    "        for i in range(500):\n",
    "            print('Collecting image {}'.format(i))\n",
    "            ret, frame = cap.read()\n",
    "            imageName = os.path.join(IMAGES_PATH, name, name+'.'+'{}.jpg'.format(str(uuid.uuid1())))\n",
    "            cv2.imwrite(imageName, frame)\n",
    "            cv2.imshow('frame', frame)\n",
    "            \n",
    "            if cv2.waitKey(1) & 0xFF == ord('q'):\n",
    "                break\n",
    "cap.release()\n",
    "cv2.destroyAllWindows()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Labelling the colllected images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!pip3 install --upgrade pyqt5 lxml"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "LABELING_PATH = os.path.join('Tensorflow', 'labelimg')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "if not os.path.exists(LABELING_PATH):\n",
    "    !mkdir {LABELING_PATH}\n",
    "    !git clone https://github.com/tzutalin/LabelImg {LABELING_PATH}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "if os.name == 'posix':\n",
    "    !cd {LABELING_PATH} && make qt5py3\n",
    "if os.name == 'nt':\n",
    "    !cd {LABELING_PATH} && pyrcc5 -o libs/resources.py resources.qrc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!cd {LABELING_PATH} && python labelImg.py"
   ]
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "f2af9fbfe845914139492b07f6927bcdd8746a3dfb5f7088a4ef2eacdfc67aa9"
  },
  "kernelspec": {
   "display_name": "Python 3.8.8 ('base')",
   "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": "3.8.13"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}