mini_cv.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "colab": {
      "name": "mini_cv.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZIua3LeV0VEC",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import numpy as np\n",
        "\n",
        "%tensorflow_version 1.x\n",
        "from keras import applications\n",
        "from keras.layers import Dense, Dropout, MaxPool2D, GlobalAveragePooling2D, Flatten\n",
        "from keras.models import Model\n",
        "from keras.optimizers import Adam, SGD\n",
        "from sklearn.metrics import confusion_matrix, roc_curve\n",
        "\n",
        "import seaborn\n",
        "\n",
        "from matplotlib import pyplot as plt\n",
        "\n",
        "import random\n",
        "from random import shuffle"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1xX7IrO3T690",
        "colab_type": "code",
        "outputId": "fc5e6e04-9a27-4873-8f37-7dfcaefb6f56",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 122
        }
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": 2,
      "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": "WEKE8PuLU0c8",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "HEADER = 'drive/My Drive/Colab Notebooks/BDRAD/mini-cv/processed/'\n",
        "\n",
        "TRAIN_EXAMPLES_PATH = HEADER + 'Mass-Training-Examples.npz'\n",
        "TRAIN_LABELS_PATH = HEADER + 'Mass-Training-Labels.npz'\n",
        "\n",
        "TEST_EXAMPLES_PATH = HEADER + 'Mass-Test-Examples.npz'\n",
        "TEST_LABELS_PATH = HEADER + 'Mass-Test-Labels.npz'"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1Yo2OFGL0VEN",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def load(examples_path, labels_path):\n",
        "    examples, labels = np.load(examples_path), np.load(labels_path)\n",
        "    examples, labels = examples.f.arr_0, labels.f.arr_0\n",
        "    examples = np.repeat(examples[..., np.newaxis], 3, -1)*65535\n",
        "    #examples = np.expand_dims(examples, axis=3) \n",
        "    num_labels = labels.shape[0]\n",
        "    one_hot = np.zeros([num_labels, 2])\n",
        "    for i in range(num_labels):\n",
        "        label = 0 if int(labels[i]) <= 2 else 1\n",
        "        one_hot[i][label] = 1\n",
        "    return examples, one_hot\n",
        "    "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Td0eSjIL0VEZ",
        "colab_type": "code",
        "outputId": "ca5251b6-2464-449f-aba3-aeb487cbde6d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        }
      },
      "source": [
        "train_examples, train_labels = load(TRAIN_EXAMPLES_PATH, TRAIN_LABELS_PATH)\n",
        "test_examples, test_labels = load(TEST_EXAMPLES_PATH, TEST_LABELS_PATH)\n",
        "\n",
        "# from matplotlib import pyplot as plt\n",
        "# plt.imshow(train_examples[2])\n",
        "#print(np.max(test_examples[2]))\n",
        "\n",
        "# plt.imshow(65535 * np.repeat(test_examples[..., np.newaxis], 3, -1)[2])\n",
        "\n",
        "def show(number, train=True):\n",
        "  if train:\n",
        "    dataset, labels = train_examples, train_labels\n",
        "  else:\n",
        "    dataset, labels = test_examples, test_labels\n",
        "  plt.figure()\n",
        "  print('train' if train else 'test')\n",
        "  print('dense' if labels[number][0] else 'not dense')\n",
        "  plt.imshow(dataset[number])\n",
        "\n",
        "def get_perc(train=True):\n",
        "  if train:\n",
        "    dataset, name = train_labels, 'train'\n",
        "  else:\n",
        "    dataset, name = test_labels, 'test'\n",
        "\n",
        "\n",
        "  num_dense = len([label for label in dataset if label[0]])\n",
        "  print('Percent dense {0}: {1}'.format(name, num_dense / len(train_labels) * 100))\n",
        "\n",
        "get_perc(True)\n",
        "get_perc(False)"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Percent dense train: 62.71186440677966\n",
            "Percent dense test: 26.27118644067797\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Rlo8UUSTYoVX",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#normalize data to get equal numbers dense and not dense to avoid skewing model\n",
        "examples = np.concatenate([train_examples, test_examples], axis=0)\n",
        "labels = np.concatenate([train_labels, test_labels], axis=0)\n",
        "\n",
        "dense, not_dense = [], []\n",
        "for i in range(len(labels)):\n",
        "  if labels[i][0]:\n",
        "    #dense example\n",
        "    dense.append((examples[i], labels[i]))\n",
        "  else:\n",
        "    #not dense example\n",
        "    not_dense.append((examples[i], labels[i]))\n",
        "  \n",
        "if len(dense) > len(not_dense):\n",
        "  dense = dense[:len(not_dense)]\n",
        "else:\n",
        "  not_dense = not_dense[:len(dense)]\n",
        "\n",
        "cutoff = int(len(dense) * 0.9)\n",
        "\n",
        "train = dense[:cutoff] + not_dense[:cutoff]\n",
        "test = dense[cutoff:] + not_dense[cutoff:]\n",
        "\n",
        "shuffle(train)\n",
        "shuffle(test)\n",
        "\n",
        "train_examples, train_labels = zip(*train)\n",
        "test_examples, test_labels = zip(*test)\n",
        "\n",
        "train_examples = np.array(train_examples)\n",
        "train_labels = np.array(train_labels)\n",
        "test_examples = np.array(test_examples)\n",
        "test_labels = np.array(test_labels) "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "B4Jy1yCg0VEh",
        "colab_type": "code",
        "outputId": "f271f644-68aa-49fb-9177-0ca77143d70f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 71
        }
      },
      "source": [
        "\"\"\"\n",
        "Modified from https://github.com/priya-dwivedi/Deep-Learning/blob/master/resnet_keras/Residual_Network_Keras.ipynb\n",
        "\"\"\"\n",
        "\n",
        "NUM_CLASSES = 2\n",
        "\n",
        "\"\"\"Tried varying levels of dropout ranging from 0.1-0.7, got best results with 0.25\"\"\"\n",
        "dropout = 0.25\n",
        "\n",
        "INPUT_SHAPE = (224, 224, 3)\n",
        "\n",
        "base_model = applications.resnet50.ResNet50(weights='imagenet', include_top=False, input_shape=INPUT_SHAPE)\n",
        "\n",
        "\"\"\"Tried only training last layers, but achieved significantly worse performance\"\"\"\n",
        "# for layer in base_model.layers:\n",
        "#   layer.trainable = False\n",
        "\n",
        "x = base_model.output\n",
        "#x = Flatten()(x)\n",
        "\n",
        "\"\"\"Prepare model for final classification\"\"\"\n",
        "x = GlobalAveragePooling2D()(x)\n",
        "\n",
        "\"\"\"Dropout layer to prevent overfitting\"\"\"\n",
        "x = Dropout(dropout)(x)\n",
        "\n",
        "\"\"\"Final dense layer for classification\"\"\"\n",
        "predictions = Dense(NUM_CLASSES, activation='softmax')(x)\n",
        "\n",
        "model = Model(inputs = base_model.input, outputs = predictions)\n",
        "\n",
        "\"\"\"Adam performed better than SGD\"\"\"\n",
        "adam = Adam(lr=0.0001)\n",
        "#sgd = SGD(lr=0.0001, momentum=0.9, decay=1e-6, nesterov=True)\n",
        "\n",
        "\"\"\"Tried binary_crossentropy vs categorical_crossentropy\"\"\"\n",
        "model.compile(optimizer=adam, loss='categorical_crossentropy', metrics=['accuracy'])"
      ],
      "execution_count": 41,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/keras_applications/resnet50.py:265: UserWarning: The output shape of `ResNet50(include_top=False)` has been changed since Keras 2.2.0.\n",
            "  warnings.warn('The output shape of `ResNet50(include_top=False)` '\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": true,
        "id": "rmgsA0Cj0VEx",
        "colab_type": "code",
        "outputId": "ed545fdd-6334-49ed-af80-ce98ba8251a3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 309
        }
      },
      "source": [
        "\"\"\"Tried 5, 7, 10 epochs\"\"\"\n",
        "epochs = 7\n",
        "\n",
        "\"\"\"Tried batch sizes of 8, 16, 32\"\"\"\n",
        "batch_size = 32\n",
        "\n",
        "model.fit(train_examples, train_labels, epochs=epochs, batch_size=batch_size,\n",
        "         validation_data=(test_examples, test_labels))\n",
        "\n",
        "\"\"\"\n",
        "Summary:\n",
        "I tuned a lot of parameters in an attempt to prevent overfitting as much as possible,\n",
        "but this was about the best I could get by changing the dropout, learning rate, optimizer, activation function, batch size, etc.\n",
        "I'd guess that the problems might be (1) a small dataset and (2) too big of a model, since these tend to be causes\n",
        "of overfitting. For the model, it's possible that a smaller ResNET architecture like ResNet-16 would have worked better,\n",
        "but I couldn't find an easy way to implement it in keras so I didn't think it would have been a good use of time for this\n",
        "particular mini-project.\n",
        "\"\"\""
      ],
      "execution_count": 42,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 350 samples, validate on 40 samples\n",
            "Epoch 1/7\n",
            "350/350 [==============================] - 41s 117ms/step - loss: 0.5970 - acc: 0.7257 - val_loss: 0.7029 - val_acc: 0.6500\n",
            "Epoch 2/7\n",
            "350/350 [==============================] - 5s 14ms/step - loss: 0.1179 - acc: 0.9600 - val_loss: 0.6446 - val_acc: 0.6750\n",
            "Epoch 3/7\n",
            "350/350 [==============================] - 5s 14ms/step - loss: 0.0315 - acc: 0.9943 - val_loss: 0.6824 - val_acc: 0.7750\n",
            "Epoch 4/7\n",
            "350/350 [==============================] - 5s 14ms/step - loss: 0.0103 - acc: 1.0000 - val_loss: 0.4623 - val_acc: 0.8750\n",
            "Epoch 5/7\n",
            "350/350 [==============================] - 5s 14ms/step - loss: 0.0054 - acc: 1.0000 - val_loss: 0.3196 - val_acc: 0.9250\n",
            "Epoch 6/7\n",
            "350/350 [==============================] - 5s 15ms/step - loss: 0.0017 - acc: 1.0000 - val_loss: 0.2859 - val_acc: 0.9500\n",
            "Epoch 7/7\n",
            "350/350 [==============================] - 5s 14ms/step - loss: 0.0016 - acc: 1.0000 - val_loss: 0.2541 - val_acc: 0.9500\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "\"\\nSummary:\\nI tuned a lot of parameters in an attempt to prevent overfitting as much as possible,\\nbut this was about the best I could get by changing the dropout, learning rate, optimizer, activation function, batch size, etc.\\nI'd guess that the problems might be (1) a small dataset and (2) too big of a model, since these tend to be causes\\nof overfitting. For the model, it's possible that a smaller ResNET architecture like ResNet-16 would have worked better,\\nbut I couldn't find an easy way to implement it in keras so I didn't think it would have been a good use of time for this\\nparticular mini-project.\\n\""
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 42
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ifpmdnGC0VFP",
        "colab_type": "code",
        "outputId": "50a3c5a2-a41e-4972-fdc6-ede5a8cf9646",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 228
        }
      },
      "source": [
        "F\"\"\"\n",
        "Adapted from https://stackoverflow.com/questions/46615554/how-to-display-multiple-images-in-one-figure-correctly/46616645\n",
        "\"\"\"\n",
        "\n",
        "fig=plt.figure()\n",
        "columns = 5\n",
        "rows = 2\n",
        "for i in range(1, columns*rows +1):\n",
        "    img = examples[int(random.random() * len(examples))]\n",
        "    fig.add_subplot(rows, columns, i)\n",
        "    plt.imshow(img)\n",
        "plt.show()"
      ],
      "execution_count": 43,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAADTCAYAAACPxNgCAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\niBAdjtDQhwgRIkSHIzT0IUKECNHh+H/XNPjZiBy8xAAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 10 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CJbnOBhthTg7",
        "colab_type": "code",
        "outputId": "dfedd313-3c85-4881-8beb-b45daad938f4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 286
        }
      },
      "source": [
        "predictions = model.predict(test_examples)\n",
        "seaborn.heatmap(confusion_matrix(test_labels.argmax(axis=1), predictions.argmax(axis=1)),\n",
        "                yticklabels=['Actual Not Dense', 'Actual Dense'], xticklabels=['Predicted Not Dense', 'Predicted Dense'])"
      ],
      "execution_count": 44,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.axes._subplots.AxesSubplot at 0x7f39106296a0>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 44
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWQAAAD8CAYAAABAWd66AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kGDODLwDt-ww",
        "colab_type": "code",
        "outputId": "18d38181-fb6a-41a5-dfad-933c72cbb704",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        }
      },
      "source": [
        "fpr, tpr, _ = roc_curve(test_labels[:,1],predictions[:,1])\n",
        "plt.plot(fpr,tpr)"
      ],
      "execution_count": 45,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[<matplotlib.lines.Line2D at 0x7f39104c3d30>]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 45
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAD4CAYAAAD8Zh1EAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSY0w6JLUiP8FNmvkTYYRwCoAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dLEfFoZMoQL0",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}