pollen_challenge.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "pollen_challenge.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "3FbN_PlIor-b",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 122
        },
        "outputId": "5881ff1a-6042-49dc-e6ef-e26074fbd5ef"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Go to this URL in a browser: https://accounts.google.com/o/oauth2/auth?client_id=947318989803-6bn6qk8qdgf4n4g3pfee6491hc0brc4i.apps.googleusercontent.com&redirect_uri=urn%3aietf%3awg%3aoauth%3a2.0%3aoob&response_type=code&scope=email%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdocs.test%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdrive%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdrive.photos.readonly%20https%3a%2f%2fwww.googleapis.com%2fauth%2fpeopleapi.readonly\n",
            "\n",
            "Enter your authorization code:\n",
            "··········\n",
            "Mounted at /content/drive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "D7Cwcbnnisxq",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "cec287f1-dffb-420d-9f01-db4928823a6f"
      },
      "source": [
        "import tensorflow as tf\n",
        "tf.__version__"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'2.2.0'"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jJ8Y5jXQjuQT",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from tensorflow.keras.models import Sequential, Model\n",
        "from tensorflow.keras.layers import Input, Dense, Conv2D, Flatten, Dropout, MaxPooling2D, Reshape, Conv1D\n",
        "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
        "from tensorflow.keras.metrics import Precision, Recall, Accuracy\n",
        "from tensorflow.keras import optimizers, layers, regularizers\n",
        "from sklearn.metrics import classification_report\n",
        "AUTOTUNE = tf.data.experimental.AUTOTUNE\n",
        "\n",
        "\n",
        "import os\n",
        "import cv2\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5NNF2f2RixMG",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# pth = '/content/drive/My Drive/Colab Notebooks/pollen_data/train/images/'"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WaiyBk_zmf0B",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# os.listdir(pth)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JCtNi8LimLB7",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# data = []\n",
        "\n",
        "# for folder in os.listdir(pth):\n",
        "#   print(folder)\n",
        "#   count = 0\n",
        "#   for subfolder in os.listdir(pth+folder):\n",
        "#     print(subfolder)\n",
        "#     if subfolder=='train_OBJ':\n",
        "#       one_hot_label = np.zeros(4)\n",
        "#       one_hot_label[int(folder)-1] = 1\n",
        "#       for image in os.listdir(pth+folder+'/'+subfolder):\n",
        "#         count += 1\n",
        "#         if count % 100 == 0 : print(count) \n",
        "#         img_matrix = cv2.imread(pth+folder+'/'+subfolder+'/'+image)\n",
        "#         data.append([img_matrix, one_hot_label])\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3FuGcnTO7CLr",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# np.save('data.npy', data)\n",
        "data = np.load('/content/drive/My Drive/Colab Notebooks/data.npy', allow_pickle=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zeP8wWg_7Ozn",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "d591edfe-5105-40c8-c604-7ed89f1fc57b"
      },
      "source": [
        "np.shape(data)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(4808, 2)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 16
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "g5w1v0og9gPq",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# count = 0\n",
        "# train_dataset = []\n",
        "# test_dataset = []\n",
        "\n",
        "# for d in data:\n",
        "#   count+=1\n",
        "#   if count<4000:\n",
        "#     train_dataset.append(d)\n",
        "#   else:\n",
        "#     test_dataset.append(d)\n",
        "  \n",
        "# print(np.shape(train_dataset), np.shape(test_dataset))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4Ne3lp2wnBDT",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 119
        },
        "outputId": "e8bf405f-0c72-4e27-957a-1393b596922c"
      },
      "source": [
        "c = np.zeros(4)\n",
        "for d in data:\n",
        "  c += d[1]\n",
        "print('Number of images from every class-',c)\n",
        "print(np.sum(c))\n",
        "\n",
        "print(np.shape(data))\n",
        "print(np.shape(data[0]))\n",
        "print(np.shape(data[0][0]))\n",
        "print(np.shape(data[0][1]))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Number of images from every class- [1566.  773. 1745.  724.]\n",
            "4808.0\n",
            "(4808, 2)\n",
            "(2,)\n",
            "(84, 84, 3)\n",
            "(4,)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Xn3e4gdR-5Xw",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "ad303965-e65a-4aa0-f387-9f32b8dee3c5"
      },
      "source": [
        "len(data)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "4808"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 19
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XE3_hIvf48T8",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "78c634f0-8343-45f2-e71e-db73b50c106c"
      },
      "source": [
        "# Data augmentation\n",
        "\n",
        "aug_data = data\n",
        "temp = []\n",
        "\n",
        "for i in range(len(data)):\n",
        "  \n",
        "  image = data[i][0]\n",
        "  label = data[i][1]\n",
        "\n",
        "  flip_image = np.fliplr(image)\n",
        "  temp = np.transpose(np.asarray([[flip_image], [label]]))\n",
        "  aug_data = np.concatenate((aug_data, temp))\n",
        "\n",
        "  flip_image2 = np.flipud(image)\n",
        "  temp = np.transpose(np.asarray([[flip_image2], [label]]))\n",
        "  aug_data = np.concatenate((aug_data, temp))\n",
        "\n",
        "  # flip_image = tf.image.flip_left_right(image)\n",
        "  # temp = np.transpose(np.asarray([[flip_image], [label]]))\n",
        "  # aug_data = np.concatenate((aug_data, temp))\n",
        "\n",
        "  # flip_image2 = tf.image.flip_up_down(image)\n",
        "  # temp = np.transpose(np.asarray([[flip_image2], [label]]))\n",
        "  # aug_data = np.concatenate((aug_data, temp))\n",
        "\n",
        "  # cropped = tf.image.central_crop(image, central_fraction=0.8)\n",
        "  # temp = [cropped, label]\n",
        "  # aug_data.append(temp)\n",
        "\n",
        "print(np.shape(aug_data), np.shape(data))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(14424, 2) (4808, 2)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "V023dz4j_rxU",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 85
        },
        "outputId": "043e1d32-56bb-471a-e2e6-50a399ba8d9c"
      },
      "source": [
        "c = np.zeros(4)\n",
        "for d in data:\n",
        "  c += d[1]\n",
        "print('Number of images from every class-',c)\n",
        "print(np.sum(c))\n",
        "\n",
        "c = np.zeros(4)\n",
        "for d in aug_data:\n",
        "  c += d[1]\n",
        "print('Number of images from every class-',c)\n",
        "print(np.sum(c))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Number of images from every class- [1566.  773. 1745.  724.]\n",
            "4808.0\n",
            "Number of images from every class- [4698. 2319. 5235. 2172.]\n",
            "14424.0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "di8-zKUU25w4",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "8c4a9a51-93a7-405e-a74a-8f42d3552afb"
      },
      "source": [
        "# shuffle\n",
        "\n",
        "np.random.seed(1)\n",
        "np.random.shuffle(aug_data)\n",
        "\n",
        "print(np.shape(aug_data))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(14424, 2)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "srDSIpouhaKr",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# X and t from data\n",
        "\n",
        "train_X = []\n",
        "train_t = []\n",
        "test_X = []\n",
        "test_t = []\n",
        "counter=0\n",
        "\n",
        "for images, labels in aug_data:\n",
        "  counter+=1\n",
        "  if counter < 12000:\n",
        "    train_X.append(images)\n",
        "    train_t.append(labels)\n",
        "  else:\n",
        "    test_X.append(images)\n",
        "    test_t.append(labels)\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QijlNvetFCUF",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "95e96975-1bd7-4d97-8fbc-0d084377fd65"
      },
      "source": [
        "print(np.shape(train_X), np.shape(train_t), np.shape(test_X), np.shape(test_t))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(11999, 84, 84, 3) (11999, 4) (2425, 84, 84, 3) (2425, 4)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UxWyvnZuiAWq",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "5c12dfa3-12e9-418c-cb74-b55394bad88a"
      },
      "source": [
        "# Normalizing the images\n",
        "\n",
        "train_X = np.asarray(train_X)/255\n",
        "train_t = np.asarray(train_t)\n",
        "\n",
        "test_X = np.asarray(test_X)/255\n",
        "test_t = np.asarray(test_t)\n",
        "\n",
        "print(np.shape(train_X), np.shape(train_t), np.shape(test_X), np.shape(test_t))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(11999, 84, 84, 3) (11999, 4) (2425, 84, 84, 3) (2425, 4)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EKbd89QziMdf",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZTzOcWQtiMwB",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Baseline CNN model\n",
        "\n",
        "input_layer = Input(shape=(84,84,3), name='input')\n",
        "conv1 = Conv2D(16, kernel_size=5, activation='relu', name='conv1', kernel_regularizer=regularizers.l2(0.001))(input_layer)\n",
        "pool1 = MaxPooling2D(pool_size=(2,2), name='pool1')(conv1)\n",
        "conv2 = Conv2D(32, kernel_size=5, activation='relu', name='conv2', kernel_regularizer=regularizers.l2(0.001))(pool1)\n",
        "pool2 = MaxPooling2D(pool_size=(2,2), name='pool2')(conv2)\n",
        "conv3 = Conv2D(64, kernel_size=5, activation='relu', name='conv3', kernel_regularizer=regularizers.l2(0.001))(pool2)\n",
        "pool3 = MaxPooling2D(pool_size=(2,2), name='pool3')(conv3)\n",
        "conv4 = Conv2D(128, kernel_size=5, activation='relu', name='conv4', kernel_regularizer=regularizers.l2(0.001))(pool3)\n",
        "flat = Flatten()(conv4)\n",
        "fc1 = Dense(100, activation='relu', name='fc1', kernel_regularizer=regularizers.l2(0.01))(flat)\n",
        "fc2 = Dense(20, activation='relu', name='fc2', kernel_regularizer=regularizers.l2(0.01))(fc1)\n",
        "fc3 = Dense(4, activation='softmax', name='fc3', kernel_regularizer=regularizers.l2(0.0001))(fc2)\n",
        "\n",
        "model1 = Model(inputs=input_layer, outputs=fc3, name='Baseline_CNN_model')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EUpOTGLAxtC_",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 578
        },
        "outputId": "a36047dd-e6c7-4de5-c862-3721269e6fcb"
      },
      "source": [
        "print(model1.summary())"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"Baseline_CNN_model\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input (InputLayer)           [(None, 84, 84, 3)]       0         \n",
            "_________________________________________________________________\n",
            "conv1 (Conv2D)               (None, 80, 80, 16)        1216      \n",
            "_________________________________________________________________\n",
            "pool1 (MaxPooling2D)         (None, 40, 40, 16)        0         \n",
            "_________________________________________________________________\n",
            "conv2 (Conv2D)               (None, 36, 36, 32)        12832     \n",
            "_________________________________________________________________\n",
            "pool2 (MaxPooling2D)         (None, 18, 18, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv3 (Conv2D)               (None, 14, 14, 64)        51264     \n",
            "_________________________________________________________________\n",
            "pool3 (MaxPooling2D)         (None, 7, 7, 64)          0         \n",
            "_________________________________________________________________\n",
            "conv4 (Conv2D)               (None, 3, 3, 128)         204928    \n",
            "_________________________________________________________________\n",
            "flatten_2 (Flatten)          (None, 1152)              0         \n",
            "_________________________________________________________________\n",
            "fc1 (Dense)                  (None, 100)               115300    \n",
            "_________________________________________________________________\n",
            "fc2 (Dense)                  (None, 20)                2020      \n",
            "_________________________________________________________________\n",
            "fc3 (Dense)                  (None, 4)                 84        \n",
            "=================================================================\n",
            "Total params: 387,644\n",
            "Trainable params: 387,644\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "None\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MKLcpJVOyj3l",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# adam = optimizers.Adam(learning_rate=0.0005)\n",
        "model1.compile(optimizer='adam', loss='categorical_crossentropy', metrics=[\"accuracy\"])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gumf3D2i-VgX",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "beb29f5f-6991-4634-b920-ceb95c2de537"
      },
      "source": [
        "m1 = model1.fit(train_X, train_t, batch_size=100, epochs=50, validation_split=0.2, shuffle=True)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/50\n",
            "96/96 [==============================] - 2s 23ms/step - loss: 1.8782 - accuracy: 0.4878 - val_loss: 1.1107 - val_accuracy: 0.6621\n",
            "Epoch 2/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.9410 - accuracy: 0.7013 - val_loss: 0.9068 - val_accuracy: 0.6850\n",
            "Epoch 3/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.7587 - accuracy: 0.7687 - val_loss: 0.6461 - val_accuracy: 0.8158\n",
            "Epoch 4/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.6738 - accuracy: 0.8020 - val_loss: 0.6005 - val_accuracy: 0.8379\n",
            "Epoch 5/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.5746 - accuracy: 0.8421 - val_loss: 0.5433 - val_accuracy: 0.8692\n",
            "Epoch 6/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.5372 - accuracy: 0.8628 - val_loss: 0.4986 - val_accuracy: 0.8796\n",
            "Epoch 7/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.5028 - accuracy: 0.8735 - val_loss: 0.5214 - val_accuracy: 0.8796\n",
            "Epoch 8/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.4766 - accuracy: 0.8827 - val_loss: 0.5280 - val_accuracy: 0.8579\n",
            "Epoch 9/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.4570 - accuracy: 0.8895 - val_loss: 0.5469 - val_accuracy: 0.8533\n",
            "Epoch 10/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.4496 - accuracy: 0.8946 - val_loss: 0.5102 - val_accuracy: 0.8767\n",
            "Epoch 11/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.4512 - accuracy: 0.8927 - val_loss: 0.5117 - val_accuracy: 0.8758\n",
            "Epoch 12/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.4298 - accuracy: 0.8994 - val_loss: 0.4968 - val_accuracy: 0.8779\n",
            "Epoch 13/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.4102 - accuracy: 0.9106 - val_loss: 0.4386 - val_accuracy: 0.8983\n",
            "Epoch 14/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.3899 - accuracy: 0.9136 - val_loss: 0.4465 - val_accuracy: 0.8979\n",
            "Epoch 15/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.3960 - accuracy: 0.9114 - val_loss: 0.4684 - val_accuracy: 0.8871\n",
            "Epoch 16/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.3836 - accuracy: 0.9174 - val_loss: 0.4482 - val_accuracy: 0.8958\n",
            "Epoch 17/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.3693 - accuracy: 0.9247 - val_loss: 0.4642 - val_accuracy: 0.8946\n",
            "Epoch 18/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.3687 - accuracy: 0.9275 - val_loss: 0.4535 - val_accuracy: 0.8983\n",
            "Epoch 19/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.3689 - accuracy: 0.9260 - val_loss: 0.4528 - val_accuracy: 0.8992\n",
            "Epoch 20/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.3603 - accuracy: 0.9301 - val_loss: 0.4562 - val_accuracy: 0.8983\n",
            "Epoch 21/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.3456 - accuracy: 0.9356 - val_loss: 0.4860 - val_accuracy: 0.8892\n",
            "Epoch 22/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.3411 - accuracy: 0.9385 - val_loss: 0.5481 - val_accuracy: 0.8687\n",
            "Epoch 23/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.3404 - accuracy: 0.9394 - val_loss: 0.5009 - val_accuracy: 0.8854\n",
            "Epoch 24/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.3311 - accuracy: 0.9420 - val_loss: 0.5026 - val_accuracy: 0.8846\n",
            "Epoch 25/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.3071 - accuracy: 0.9529 - val_loss: 0.4936 - val_accuracy: 0.8908\n",
            "Epoch 26/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.3228 - accuracy: 0.9458 - val_loss: 0.4928 - val_accuracy: 0.8988\n",
            "Epoch 27/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.2925 - accuracy: 0.9575 - val_loss: 0.5219 - val_accuracy: 0.8854\n",
            "Epoch 28/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.2976 - accuracy: 0.9555 - val_loss: 0.4954 - val_accuracy: 0.8979\n",
            "Epoch 29/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.3062 - accuracy: 0.9543 - val_loss: 0.5584 - val_accuracy: 0.8583\n",
            "Epoch 30/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.3074 - accuracy: 0.9538 - val_loss: 0.5256 - val_accuracy: 0.8850\n",
            "Epoch 31/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.2732 - accuracy: 0.9685 - val_loss: 0.5550 - val_accuracy: 0.8725\n",
            "Epoch 32/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.2538 - accuracy: 0.9730 - val_loss: 0.5236 - val_accuracy: 0.8917\n",
            "Epoch 33/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.2701 - accuracy: 0.9669 - val_loss: 0.5499 - val_accuracy: 0.8900\n",
            "Epoch 34/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.2719 - accuracy: 0.9654 - val_loss: 0.5517 - val_accuracy: 0.8833\n",
            "Epoch 35/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.2573 - accuracy: 0.9720 - val_loss: 0.5505 - val_accuracy: 0.8846\n",
            "Epoch 36/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.2512 - accuracy: 0.9724 - val_loss: 0.5672 - val_accuracy: 0.8854\n",
            "Epoch 37/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.2622 - accuracy: 0.9695 - val_loss: 0.5735 - val_accuracy: 0.8846\n",
            "Epoch 38/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.2668 - accuracy: 0.9654 - val_loss: 0.6123 - val_accuracy: 0.8733\n",
            "Epoch 39/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.2463 - accuracy: 0.9769 - val_loss: 0.5565 - val_accuracy: 0.8913\n",
            "Epoch 40/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.2382 - accuracy: 0.9767 - val_loss: 0.5803 - val_accuracy: 0.8963\n",
            "Epoch 41/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.2378 - accuracy: 0.9790 - val_loss: 0.5856 - val_accuracy: 0.8871\n",
            "Epoch 42/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.2276 - accuracy: 0.9822 - val_loss: 0.5802 - val_accuracy: 0.8879\n",
            "Epoch 43/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.2153 - accuracy: 0.9874 - val_loss: 0.5959 - val_accuracy: 0.8892\n",
            "Epoch 44/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.2310 - accuracy: 0.9812 - val_loss: 0.6027 - val_accuracy: 0.8767\n",
            "Epoch 45/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.2126 - accuracy: 0.9881 - val_loss: 0.6180 - val_accuracy: 0.8800\n",
            "Epoch 46/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.2586 - accuracy: 0.9697 - val_loss: 0.5541 - val_accuracy: 0.8821\n",
            "Epoch 47/50\n",
            "96/96 [==============================] - 2s 20ms/step - loss: 0.2156 - accuracy: 0.9865 - val_loss: 0.6076 - val_accuracy: 0.8908\n",
            "Epoch 48/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.2225 - accuracy: 0.9822 - val_loss: 0.6341 - val_accuracy: 0.8729\n",
            "Epoch 49/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.2130 - accuracy: 0.9869 - val_loss: 0.6273 - val_accuracy: 0.8729\n",
            "Epoch 50/50\n",
            "96/96 [==============================] - 2s 21ms/step - loss: 0.1917 - accuracy: 0.9933 - val_loss: 0.5924 - val_accuracy: 0.8908\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "FB4etX1s_Z4l",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "test_y = model1.predict(test_X)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nP5IQxOC51Ps",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "y_true = np.array(tf.argmax(test_t, 1))\n",
        "y_pred = np.array(tf.argmax(test_y, 1))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_7z-rN438skC",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "d1e6f9ae-20ca-42e1-c80a-4862838958ff"
      },
      "source": [
        "print(classification_report(y_true, y_pred, digits=4))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0     0.8835    0.8698    0.8766       776\n",
            "           1     0.8938    0.8498    0.8712       406\n",
            "           2     0.8632    0.9018    0.8821       896\n",
            "           3     0.9027    0.8818    0.8921       347\n",
            "\n",
            "    accuracy                         0.8800      2425\n",
            "   macro avg     0.8858    0.8758    0.8805      2425\n",
            "weighted avg     0.8805    0.8800    0.8800      2425\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RCWGIw2YhZ6D",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 573
        },
        "outputId": "f2d337ae-dd36-4abd-e3dc-d30722b318b6"
      },
      "source": [
        "plt.plot(m1.history['accuracy'])\n",
        "plt.plot(m1.history['val_accuracy'])\n",
        "plt.title('model1 accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'validation'], loc='upper left')\n",
        "plt.grid()\n",
        "plt.show()\n",
        "\n",
        "plt.plot(m1.history['loss'])\n",
        "plt.plot(m1.history['val_loss'])\n",
        "plt.title('model1 loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'validation'], loc='upper right')\n",
        "plt.grid()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "v5fyPkCjPWx6",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7SL77p-zQFR6",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "N2Mc2hysQFe2",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Global reasoning module on the baseline model\n",
        "\n",
        "input_layer = Input(shape=(84,84,3), name='input')\n",
        "conv1 = Conv2D(16, kernel_size=5, activation='relu', name='conv1', kernel_regularizer=regularizers.l2(0.0005))(input_layer)\n",
        "\n",
        "# Graph-based global reasoning module # input is 80x80x16\n",
        "graph_transpose = Reshape((16, 80*80))(conv1)\n",
        "squeezed_graph_transpose = Conv1D(100, 1, activation='relu', name='squeezer', kernel_regularizer=regularizers.l2(0.0001))(graph_transpose)\n",
        "squeezed_graph = Reshape((100, 16))(squeezed_graph_transpose)\n",
        "gconv = Conv1D(8, 1, activation='relu', name='gconv', kernel_regularizer=regularizers.l2(0.0001))(squeezed_graph)\n",
        "gconv_transpose = Reshape((8, 100))(gconv)\n",
        "unsqueezed_graph_transpose = Conv1D(6400, 1, activation='relu', name='unsqueezer', kernel_regularizer=regularizers.l2(0.0001))(gconv_transpose)\n",
        "unsqueezed_graph = Reshape((6400, 8))(unsqueezed_graph_transpose)\n",
        "glore_image = Reshape((80, 80, 8))(unsqueezed_graph)\n",
        "\n",
        "block1 = layers.concatenate([conv1, glore_image], axis=-1)\n",
        "pool1 = MaxPooling2D(pool_size=(2,2), name='pool1')(block1)\n",
        "conv2 = Conv2D(32, kernel_size=5, activation='relu', name='conv2', kernel_regularizer=regularizers.l2(0.0005))(pool1)\n",
        "\n",
        "# Graph-based global reasoning module 2 # input is 36x36x32\n",
        "graph_transpose2 = Reshape((32, 36*36))(conv2)\n",
        "squeezed_graph_transpose2 = Conv1D(100, 1, activation='relu', name='squeezer2', kernel_regularizer=regularizers.l2(0.0001))(graph_transpose2)\n",
        "squeezed_graph2 = Reshape((100, 32))(squeezed_graph_transpose2)\n",
        "gconv2 = Conv1D(16, 1, activation='relu', name='gconv2', kernel_regularizer=regularizers.l2(0.0001))(squeezed_graph2)\n",
        "gconv_transpose2 = Reshape((16, 100))(gconv2)\n",
        "unsqueezed_graph_transpose2 = Conv1D(36*36, 1, activation='relu', name='unsqueezer2', kernel_regularizer=regularizers.l2(0.0001))(gconv_transpose2)\n",
        "unsqueezed_graph2 = Reshape((36*36, 16))(unsqueezed_graph_transpose2)\n",
        "glore_image2 = Reshape((36, 36, 16))(unsqueezed_graph2)\n",
        "\n",
        "block2 = layers.concatenate([conv2, glore_image2], axis=-1)\n",
        "pool2 = MaxPooling2D(pool_size=(2,2), name='pool2')(block2)\n",
        "conv3 = Conv2D(64, kernel_size=5, activation='relu', name='conv3', kernel_regularizer=regularizers.l2(0.0005))(pool2)\n",
        "pool3 = MaxPooling2D(pool_size=(2,2), name='pool3')(conv3)\n",
        "conv4 = Conv2D(128, kernel_size=5, activation='relu', name='conv4', kernel_regularizer=regularizers.l2(0.0005))(pool3)\n",
        "flat = Flatten()(conv4)\n",
        "fc1 = Dense(100, activation='relu', name='fc1', kernel_regularizer=regularizers.l2(0.01))(flat)\n",
        "fc2 = Dense(20, activation='relu', name='fc2', kernel_regularizer=regularizers.l2(0.01))(fc1)\n",
        "fc3 = Dense(4, activation='softmax', name='fc3', kernel_regularizer=regularizers.l2(0.0001))(fc2)\n",
        "\n",
        "model2 = Model(inputs=input_layer, outputs=fc3, name='CNN_model_with_global_reasoning_module')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2-27sR8gSEjB",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# print(model2.summary())"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dpR97B8ln6vB",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# tf.keras.utils.plot_model(model2, 'a.png')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7sGsh07DhO9N",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "adam = optimizers.Adam(learning_rate=0.0001)\n",
        "model2.compile(optimizer='adam', loss='categorical_crossentropy', metrics=[\"accuracy\"])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PCt6N79ixLeS",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 697
        },
        "outputId": "6e998208-acae-4774-8cb1-3a4c7f469a2a"
      },
      "source": [
        "m2 = model2.fit(train_X, train_t, batch_size=100, epochs=20, validation_split=0.2)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "96/96 [==============================] - 4s 38ms/step - loss: 1.9872 - accuracy: 0.4183 - val_loss: 1.3149 - val_accuracy: 0.5871\n",
            "Epoch 2/20\n",
            "96/96 [==============================] - 3s 34ms/step - loss: 1.0594 - accuracy: 0.6447 - val_loss: 0.8989 - val_accuracy: 0.6954\n",
            "Epoch 3/20\n",
            "96/96 [==============================] - 3s 33ms/step - loss: 0.8569 - accuracy: 0.7047 - val_loss: 0.7807 - val_accuracy: 0.7308\n",
            "Epoch 4/20\n",
            "96/96 [==============================] - 3s 33ms/step - loss: 0.7258 - accuracy: 0.7654 - val_loss: 0.7351 - val_accuracy: 0.7525\n",
            "Epoch 5/20\n",
            "96/96 [==============================] - 3s 34ms/step - loss: 0.6404 - accuracy: 0.8034 - val_loss: 0.5870 - val_accuracy: 0.8283\n",
            "Epoch 6/20\n",
            "96/96 [==============================] - 3s 33ms/step - loss: 0.5393 - accuracy: 0.8493 - val_loss: 0.5270 - val_accuracy: 0.8508\n",
            "Epoch 7/20\n",
            "96/96 [==============================] - 3s 34ms/step - loss: 0.5076 - accuracy: 0.8600 - val_loss: 0.6155 - val_accuracy: 0.8037\n",
            "Epoch 8/20\n",
            "96/96 [==============================] - 3s 33ms/step - loss: 0.4442 - accuracy: 0.8818 - val_loss: 0.5069 - val_accuracy: 0.8633\n",
            "Epoch 9/20\n",
            "96/96 [==============================] - 3s 33ms/step - loss: 0.4124 - accuracy: 0.8970 - val_loss: 0.4910 - val_accuracy: 0.8679\n",
            "Epoch 10/20\n",
            "96/96 [==============================] - 3s 33ms/step - loss: 0.4042 - accuracy: 0.8996 - val_loss: 0.4468 - val_accuracy: 0.8858\n",
            "Epoch 11/20\n",
            "96/96 [==============================] - 3s 33ms/step - loss: 0.3742 - accuracy: 0.9124 - val_loss: 0.5217 - val_accuracy: 0.8596\n",
            "Epoch 12/20\n",
            "96/96 [==============================] - 3s 34ms/step - loss: 0.3323 - accuracy: 0.9307 - val_loss: 0.4598 - val_accuracy: 0.8871\n",
            "Epoch 13/20\n",
            "96/96 [==============================] - 3s 33ms/step - loss: 0.3249 - accuracy: 0.9355 - val_loss: 0.4916 - val_accuracy: 0.8858\n",
            "Epoch 14/20\n",
            "96/96 [==============================] - 3s 33ms/step - loss: 0.2954 - accuracy: 0.9434 - val_loss: 0.5662 - val_accuracy: 0.8550\n",
            "Epoch 15/20\n",
            "96/96 [==============================] - 3s 33ms/step - loss: 0.2731 - accuracy: 0.9558 - val_loss: 0.5948 - val_accuracy: 0.8637\n",
            "Epoch 16/20\n",
            "96/96 [==============================] - 3s 33ms/step - loss: 0.2471 - accuracy: 0.9655 - val_loss: 0.5462 - val_accuracy: 0.8804\n",
            "Epoch 17/20\n",
            "96/96 [==============================] - 3s 33ms/step - loss: 0.2723 - accuracy: 0.9553 - val_loss: 0.5299 - val_accuracy: 0.8825\n",
            "Epoch 18/20\n",
            "96/96 [==============================] - 3s 33ms/step - loss: 0.2281 - accuracy: 0.9729 - val_loss: 0.5793 - val_accuracy: 0.8825\n",
            "Epoch 19/20\n",
            "96/96 [==============================] - 3s 33ms/step - loss: 0.2256 - accuracy: 0.9715 - val_loss: 0.6590 - val_accuracy: 0.8575\n",
            "Epoch 20/20\n",
            "96/96 [==============================] - 3s 33ms/step - loss: 0.2096 - accuracy: 0.9775 - val_loss: 0.6032 - val_accuracy: 0.8833\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lQj86GpQxUDg",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "test_y = model2.predict(test_X)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uwwQQ5nfxvXC",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "y_true = np.array(tf.argmax(test_t, 1))\n",
        "y_pred = np.array(tf.argmax(test_y, 1))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6VPI-66Wx2Xk",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "58596e37-ca46-4541-8e29-68d562dba46f"
      },
      "source": [
        "print(classification_report(y_true, y_pred, digits=4))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0     0.8564    0.8686    0.8624       776\n",
            "           1     0.8622    0.8325    0.8471       406\n",
            "           2     0.8543    0.8571    0.8557       896\n",
            "           3     0.9020    0.9020    0.9020       347\n",
            "\n",
            "    accuracy                         0.8631      2425\n",
            "   macro avg     0.8687    0.8651    0.8668      2425\n",
            "weighted avg     0.8631    0.8631    0.8631      2425\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KdKm4zP8x6Rh",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 573
        },
        "outputId": "c4a086bb-d375-4ff9-e25b-f8c260611fac"
      },
      "source": [
        "plt.plot(m2.history['accuracy'])\n",
        "plt.plot(m2.history['val_accuracy'])\n",
        "plt.title('model2 accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'validation'], loc='upper left')\n",
        "plt.grid()\n",
        "plt.show()\n",
        "\n",
        "plt.plot(m2.history['loss'])\n",
        "plt.plot(m2.history['val_loss'])\n",
        "plt.title('model2 loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'validation'], loc='upper right')\n",
        "plt.grid()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "364NnpN-giPr",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}