Deep Learning Challenge Auto Tag Images Gala.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Hackerearth Deep Learning Challenge Auto Tag Images Gala.ipynb",
      "provenance": [],
      "authorship_tag": "ABX9TyMJ+pg63Up+jD5stNhOKp4O",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/ArpanSM/Machine_Learning_Hackathons/blob/master/Deep%20Learning%20Challenge%20Auto%20Tag%20Images%20Gala.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "h8Z1MbUsWRJZ"
      },
      "source": [
        "import tensorflow as tf\r\n",
        "from tensorflow.keras.models import Sequential\r\n",
        "from tensorflow.keras.layers import Dense, Convolution2D, MaxPooling2D, Flatten,Conv2D,BatchNormalization,Dropout\r\n",
        "import numpy as np\r\n",
        "import matplotlib.pyplot as plt\r\n",
        "import pandas as pd\r\n",
        "#print(tf.__version__)\r\n",
        "\r\n",
        "#GPU Training\r\n",
        "import keras\r\n",
        "import tensorflow as tf\r\n",
        "\r\n",
        "\r\n",
        "config = tf.ConfigProto( device_count = {'GPU': 1 , 'CPU': 4} ) \r\n",
        "sess = tf.Session(config=config) \r\n",
        "keras.backend.set_session(sess)\r\n",
        "\r\n",
        "\r\n",
        "\r\n",
        "training_set = pd.read_csv(\"./train.csv\")\r\n",
        "training_imgs = [\"{}\".format(x) for x in list(training_set.Image)]\r\n",
        "training_label= list(training_set['Class'])\r\n",
        "training_set = pd.DataFrame( {'Images': training_imgs,'Class': training_label} )\r\n",
        "#Changing the type  to str\r\n",
        "training_set.Class = training_set.Class.astype(str)\r\n",
        "#print(training_set.head())\r\n",
        "\r\n",
        "#data-preprocessing\r\n",
        "from tensorflow.keras.preprocessing.image import ImageDataGenerator\r\n",
        "train_dataGen = ImageDataGenerator(rescale = 1./255,\r\n",
        "                                  shear_range = 0.2,\r\n",
        "                                  zoom_range = 0.2,\r\n",
        "                                  horizontal_flip = True)\r\n",
        "train_generator = train_dataGen.flow_from_dataframe(\r\n",
        "                                        dataframe = training_set,\r\n",
        "                                        directory=\"./Train\",\r\n",
        "                                        x_col=\"Images\",\r\n",
        "                                        y_col=\"Class\",\r\n",
        "                                        class_mode=\"categorical\",\r\n",
        "                                        target_size=(128,128),\r\n",
        "                                        batch_size=32)\r\n",
        "\r\n",
        "input_shape = (128,128,3)\r\n",
        "num_classes = 4\r\n",
        "\r\n",
        "#Building Convolutional Neural Network\r\n",
        "classifier = Sequential()\r\n",
        "#First Convolutional layer\r\n",
        "classifier.add(Convolution2D(filters = 64,kernel_size = (3,3), activation = 'relu', input_shape = (128,128,3)))\r\n",
        "classifier.add(MaxPooling2D(pool_size = (2,2)))\r\n",
        "#second Convolutional layer\r\n",
        "classifier.add(Convolution2D(128,(3,3),activation = 'relu'))\r\n",
        "classifier.add(MaxPooling2D(pool_size = (2,2)))\r\n",
        "#third\r\n",
        "classifier.add(Convolution2D(256,(3,3),activation = 'relu'))\r\n",
        "classifier.add(MaxPooling2D(pool_size = (2,2)))\r\n",
        "#Flattening\r\n",
        "classifier.add(Flatten())\r\n",
        "#Hidden Layer\r\n",
        "classifier.add(Dense(units = 64, activation = 'relu'))\r\n",
        "classifier.add(Dense(units = 128, activation = 'relu'))\r\n",
        "classifier.add(Dense(units = 128, activation = 'relu'))\r\n",
        "classifier.add(Dense(units = 256, activation = 'relu'))\r\n",
        "#Output Layer\r\n",
        "classifier.add(Dense(units = 4 , activation = 'softmax'))\r\n",
        "classifier.compile(optimizer = 'Adamax', loss = 'categorical_crossentropy', metrics = ['categorical_accuracy','accuracy'])\r\n",
        "#classifier.summary()\r\n",
        "#Training the CNN\r\n",
        "classifier.fit_generator(train_generator, epochs = 10, steps_per_epoch = 100 )\r\n",
        "\r\n",
        "\r\n",
        "\r\n",
        "#Predicting For Test Set\r\n",
        "test_set = pd.read_csv(\"C:/Users/arpan/OneDrive/Desktop/Gala_Classify/test.csv\")\r\n",
        "test_imgs = [\"./Test/{}\".format(x) for x in list(test_set.Image)]\r\n",
        "test_set = pd.DataFrame( {'Images': test_imgs })\r\n",
        "classes = train_generator.class_indices\r\n",
        "print(classes)\r\n",
        "inverted_classes = dict(map(reversed, classes.items()))\r\n",
        "print(inverted_classes)\r\n",
        "\r\n",
        "\r\n",
        "#Predicting classes\r\n",
        "from keras.preprocessing import image\r\n",
        "Y_pred = []\r\n",
        "for i in range(len(test_set)):\r\n",
        "  img = image.load_img(path= test_set.Images[i],target_size=(128,128,3))\r\n",
        "  img = image.img_to_array(img)\r\n",
        "  test_img = img.reshape((1,128,128,3))\r\n",
        "  img_class = classifier.predict_classes(test_img)\r\n",
        "  prediction = img_class[0]\r\n",
        "  Y_pred.append(prediction)\r\n",
        "#print(Y_pred)\r\n",
        "prediction_classes = [ inverted_classes.get(item,item) for item in Y_pred ]\r\n",
        "#print(prediction_classes)\r\n",
        "\r\n",
        "\r\n",
        "\r\n",
        "classifier.save_weights(\"model.h5\")\r\n",
        "print(\"Saved model to disk\")\r\n",
        "\r\n",
        "#Writing to excel\r\n",
        "#test_set['Class'] = prediction_classes\r\n",
        "#test_set.to_csv('test.csv', index=False)"
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}