{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "computer_vision.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "metadata": {
        "id": "o8rKg5jPF_aa",
        "colab_type": "code",
        "outputId": "9569d1de-a4e6-42b0-ab60-713b627ec02d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 53
        }
      },
      "cell_type": "code",
      "source": [
        "import os\n",
        "import csv\n",
        "import numpy as np\n",
        "\n",
        "repo_location = os.path.join('/content', 'e4-gan')\n",
        "print(repo_location)\n",
        "if not os.path.exists(repo_location):\n",
        "  !git clone https://git.skozl.com/e4-gan /content/e4-gan\n",
        "    \n",
        "os.chdir(repo_location)\n",
        "!cd /content/e4-gan\n",
        "!git pull\n"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/content/e4-gan\n",
            "Already up to date.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "Mci7b38-bDjf",
        "colab_type": "code",
        "outputId": "0ec49551-a260-4469-f656-146b6a3bb226",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 161
        }
      },
      "cell_type": "code",
      "source": [
        "imgfolder = os.path.join(repo_location, 'images')\n",
        "print(imgfolder)\n",
        "if not os.path.exists(imgfolder):\n",
        "  !mkdir images\n",
        "  print('Make image directory')\n",
        "  \n",
        "from dcgan import DCGAN\n",
        "from cgan import CGAN\n",
        "from cdcgan import CDCGAN\n",
        "from lenet import *\n",
        "  \n",
        "#vbn_dcgan = DCGAN(virtual_batch_normalization=True)\n",
        "#utils = os.path.join('/content', 'utils')\n",
        "cgan = CGAN()\n",
        "cdcgan = CDCGAN()\n",
        "\n",
        "#dcgan.train(epochs=4000, batch_size=32, save_interval=1000)\n",
        "#cgan.train(epochs=20000, batch_size=32, sample_interval=1000, graph=True)"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/content/e4-gan/images\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Colocations handled automatically by placer.\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "LcifrT3feO6P",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "#cdcgan.discriminator.save_weights('disc_weights.h5')\n",
        "#cdcgan.generator.save_weights('gen_weights.h5')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "X_-PBBXitdui",
        "colab_type": "code",
        "outputId": "b49313cf-54b3-44ee-9afb-3dfe1d16906d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 125
        }
      },
      "cell_type": "code",
      "source": [
        "cdcgan.train(epochs=10001, batch_size=128, sample_interval=200, graph=True, smooth_real=0.9)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\r  0%|          | 0/10001 [00:00<?, ?it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use tf.cast instead.\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/keras/engine/training.py:490: UserWarning: Discrepancy between trainable weights and collected trainable weights, did you set `model.trainable` without calling `model.compile` after ?\n",
            "  'Discrepancy between trainable weights and collected trainable'\n",
            "  7%|▋         | 728/10001 [01:38<19:14,  8.03it/s]"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "metadata": {
        "id": "a56uNnvlwZgt",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "#cgan.train(epochs=10000, batch_size=32, sample_interval=1000, graph=True, smooth_real=0.9)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "ZYR97BHmMuQE",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "X_train, y_train, X_validation, y_validation, X_test, y_test = import_mnist()\n",
        "train_gen, test_gen, val_gen, tr_labels_gen, te_labels_gen, val_labels_gen = cdcgan.generate_data()\n",
        "\n",
        "# If split = 0 use only original mnist set\n",
        "train_data, train_labels, val_data, val_labels = mix_data(X_train, y_train, X_validation, y_validation, train_gen, tr_labels_gen, val_gen, val_labels_gen, split=0.3)\n",
        "print(val_data.shape, val_labels.shape, train_data.shape, train_labels.shape)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "zbrG6Uk8Tfqd",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "cdcgan.generator.save('gen.h5')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Lfd0uuM0m98s",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "model = train_classifier(train_data, train_labels, X_validation, y_validation, batch_size=128, epochs=100)\n",
        "#For further steps of fine tuning use:\n",
        "#model.fit(train_data, train_labels, batch_size=128, epochs=100, verbose=1, validation_data = (X_validation, y_validation))\n",
        "model.save_weights('lenet.h5')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "8e-UgoZ7et9D",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "model = get_lenet_icp((32,32,1))\n",
        "model.load_weights('lenet.h5')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "CiGcNvjeNOjp",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "accuracy_mnist, inception_mnist = test_classifier(model, X_test, y_test)\n",
        "print('Accuracy', accuracy_mnist)\n",
        "print('Inception Score', inception_mnist)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "mWPYOjK3X3cS",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "accuracy_gen, inception_gen = test_classifier(model, test_gen, te_labels_gen)\n",
        "print('Accuracy', accuracy_gen)\n",
        "print('Inception Score', inception_gen)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "ZxTMGlwuj9vu",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "'''\n",
        "import matplotlib.pyplot as plt \n",
        "\n",
        "precision_mnist = np.array(11)\n",
        "inception_score = np.array(11)\n",
        "\n",
        "for i in range(11):\n",
        "  split = float(i)/10\n",
        "  train_data, train_labels, val_data, val_labels = mix_data(X_train, y_train, X_validation, y_validation, train_gen, tr_labels_gen, val_gen, val_labels_gen, split=split)\n",
        "  model = train_classifier(train_data, train_labels, X_validation, y_validation, batch_size=128, epochs=100)\n",
        "  precision_mnist[i] = test_classifier(model, X_test, y_test)\n",
        "  inception_score[i] = test_classifier(model, test_gen, te_labels_gen)\n",
        "  \n",
        "xgrid = 100*np.arange(11)\n",
        "plt.plot(xgrid, 100*precision_mnist)\n",
        "plt.plot(xgrid, 100*inception_score)\n",
        "plt.ylabel('Classification Accuracy (%)')\n",
        "plt.xlabel('Amount of generated data used for training')\n",
        "plt.legend(('MNIST Test Set', 'CGAN Generated Test Set'), loc='best')\n",
        "plt.show()\n",
        "'''"
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}