{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"import numpy as np\nimport matplotlib.pyplot as plt\nimport matplotlib as mpl\nimport pandas as pd\nmpl.rcParams['font.sans-serif'] = ['SimHei']\nmpl.rcParams['axes.unicode_minus'] = False\npd.set_option('display.max_columns', None)\npd.set_option('display.max_rows', None)\nimport tensorflow as tf \nfrom tensorflow.keras import Sequential\nfrom tensorflow.keras.layers import * \nplt.style.use('classic')\nimport os \nfrom sklearn.metrics import confusion_matrix, roc_auc_score, roc_curve, classification_report\nfrom tensorflow.keras.preprocessing import image\nfrom sklearn.model_selection import train_test_split\nfrom tensorflow.keras.optimizers import Adam\n# import cv2 ","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"select_img = []\nfor img_name in os.listdir('./train'):\n    if int(img_name.split('.')[1]) <= 500:\n        select_img.append(img_name)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"img = plt.imread('./train/cat.0.jpg')","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"choose 5001 images ","metadata":{}},{"cell_type":"code","source":"for i in range(5001):\n    img_cat = plt.imread(f'./train/cat.{i}.jpg')\n    img_dog = plt.imread(f'./train/dog.{i}.jpg')\n    img_cat = cv2.resize(img_cat, (256, 256))\n    img_dog = cv2.resize(img_dog, (256, 256))\n    plt.imsave(f'./output/cat.{i}.jpg', img_cat)\n    plt.imsave(f'./output/dog.{i}.jpg', img_dog)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"img = cv2.resize(img, (150, 150)) ","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"plt.imsave('./demo.jpg', img)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"read 2501 for cats and 2501 for dogs, they have the same shape (150, 150)","metadata":{}},{"cell_type":"code","source":"cat_lis = []\ndog_lis = []\nfor i in range(2501):\n    cat_lis.append(plt.imread(f'./output/cat.{i}.jpg').tolist())\n    dog_lis.append(plt.imread(f'./output/dog.{i}.jpg').tolist())","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"cat_lis = np.array(cat_lis)\ndog_lis = np.array(dog_lis)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"target = np.array([0 for _ in range(2501)] + [1 for _ in range(2501)])","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"feature = np.concatenate((cat_lis, dog_lis))","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"feature.shape ","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"x_train, x_test, y_train, y_test = train_test_split(feature, target, stratify=target, test_size=0.1)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## LENET5","metadata":{}},{"cell_type":"code","source":"model = Sequential()\nmodel.add(Conv2D(kernel_size=3, input_shape = (150, 150, 3), filters=16, activation=\"relu\"))\nmodel.add(MaxPool2D(pool_size=2, strides=2))\nmodel.add(Conv2D(kernel_size=3, filters=32, activation=\"relu\"))\nmodel.add(MaxPool2D(pool_size=2, strides=2))\nmodel.add(Flatten())\nmodel.add(Dense(512, activation=\"relu\"))\nmodel.add(Dropout(0.5))\nmodel.add(Dense(256, activation=\"relu\"))\nmodel.add(Dropout(0.5))\n\nmodel.add(Dense(1, activation=\"sigmoid\"))","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"adam = Adam(learning_rate=1e-4)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model.compile(loss=\"binary_crossentropy\", optimizer=adam, metrics=[\"accuracy\"])\n","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"x_train.max()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"(y_train == 1).sum()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"regulation the feature","metadata":{}},{"cell_type":"code","source":"x_train = x_train / 255\nx_test = x_test / 255\n# y_train = tf.one_hot(y_train, depth=2)\n# y_test = tf.one_hot(y_test, depth=2)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model.fit(x_train, y_train, validation_data=(x_test, y_test), epochs=100, batch_size=20)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"check loss ","metadata":{"execution":{"iopub.status.busy":"2022-03-01T06:17:57.209712Z","iopub.execute_input":"2022-03-01T06:17:57.210097Z","iopub.status.idle":"2022-03-01T06:17:57.240518Z","shell.execute_reply.started":"2022-03-01T06:17:57.210006Z","shell.execute_reply":"2022-03-01T06:17:57.239122Z"}}},{"cell_type":"code","source":"history = model.history.history\nplt.title(\"loss on both train and test\")\nplt.plot(history[\"loss\"], label = \"train\")\nplt.plot(history[\"val_loss\"], label = \"test\")\nplt.legend()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"check acuracy","metadata":{}},{"cell_type":"code","source":"plt.title(\"accuracy on both train and test\")\nplt.plot(history[\"accuracy\"], label = \"train\")\nplt.plot(history[\"val_accuracy\"], label = \"test\")\nplt.legend()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"drow roc and show confusematrix and classification report","metadata":{}},{"cell_type":"code","source":"result = model(x_test)\nx,y , _ = roc_curve(y_test, result)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"\nplt.plot(x, y)\nplt.plot(x, x,color = \"orange\", ls = \":\")\nplt.title(\"auc = \" + str(roc_auc_score(y_test, result)))","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"confusion_matrix(y_test, np.round(model(x_test)))","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(classification_report(y_test, np.round(model(x_test))))","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## AlexNet","metadata":{}},{"cell_type":"code","source":"\nmodel = Sequential()\nmodel.add(Conv2D(32, (3, 3), activation='relu', input_shape=(150, 150, 3)))\nmodel.add(MaxPooling2D((2, 2)))\nmodel.add(Conv2D(64, (3, 3), activation='relu'))\nmodel.add(MaxPooling2D((2, 2)))\nmodel.add(Conv2D(128, (3, 3), activation='relu'))\nmodel.add(MaxPooling2D((2, 2)))\nmodel.add(Conv2D(128, (3, 3), activation='relu'))\nmodel.add(MaxPooling2D((2, 2)))\nmodel.add(Flatten())\nmodel.add(Dense(512, activation='relu'))\nmodel.add(Dense(1, activation='sigmoid'))","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model.compile(loss=\"binary_crossentropy\", optimizer=\"adam\", metrics=[\"accuracy\"])\n\nmodel.fit(x_train, y_train, validation_data=(x_test, y_test), epochs=100, batch_size=20)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"history = model.history.history\nplt.title(\"loss on both train and test\")\nplt.plot(history[\"loss\"], label = \"train\")\nplt.plot(history[\"val_loss\"], label = \"test\")\nplt.legend()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"plt.title(\"accuracy on both train and test\")\nplt.plot(history[\"accuracy\"], label = \"train\")\nplt.plot(history[\"val_accuracy\"], label = \"test\")\nplt.legend()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"result = model(x_test)\nx,y , _ = roc_curve(y_test, result)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"\nplt.plot(x, y)\nplt.plot(x, x,color = \"orange\", ls = \":\")\nplt.title(\"auc = \" + str(roc_auc_score(y_test, result)))","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"confusion_matrix(y_test, np.round(model(x_test)))","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(classification_report(y_test, np.round(model(x_test))))","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Resnet","metadata":{}},{"cell_type":"code","source":"def conv_global(input_x, kernel_size, filters, padding, stride):\n    x = Conv2D(kernel_size=kernel_size, filters=filters, padding=padding, strides=stride)(input_x)\n    x = BatchNormalization()(x)\n    x = Activation(\"relu\")(x)\n    return x \n","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"x = np.random.randn(4, 255, 255, 3)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"conv_global(x, (3, 3), 16, \"same\", 1).shape ","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def channel_a(input_x, channel = 64):\n    x = conv_global(input_x, (3, 3), channel, \"same\", 1)\n    x = conv_global(x, (3, 3), channel, \"same\", 1)\n    y = x + input_x\n    return y \n","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def channel_b(input_x, channel = 128):\n    x = conv_global(input_x, (3, 3), channel, \"same\", 2)\n    x = conv_global(x, (3, 3), channel, \"same\", 1)\n\n    input_x = conv_global(input_x, (3, 3), channel, \"same\", 2)\n    y = x + input_x\n    return y   ","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"x = np.random.randn(32, 150, 150, 64)\nchannel_b(x).shape ","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"from tensorflow.keras import Model","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"input_layer = Input((150, 150, 3))\nconv1= conv_global(input_layer, (7, 7), 64, \"same\", 1)\nconv1_pooling = MaxPool2D(pool_size=(3, 3), strides=2, padding=\"same\")(conv1)\n\nx = channel_b(conv1_pooling, channel=64)\nx = channel_a(x, channel=64)\n\nx = channel_b(x, 128)\nx = channel_a(x, 128)\n\nx = channel_b(x, 256)\nx = channel_a(x, 256)\n\nx = channel_b(x, 512)\nx = channel_a(x, 512)\n\nx = GlobalAveragePooling2D()(x)\nx = Flatten()(x)\n\nx = Dense(1, activation=\"sigmoid\")(x)\n\n\nmodel = Model([input_layer], [x])","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model.summary()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model.compile(loss=\"binary_crossentropy\", optimizer=\"adam\", metrics=[\"accuracy\"])\n\nmodel.fit(x_train, y_train, validation_data=(x_test, y_test), epochs=100, batch_size=20)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"y_train.shape ","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"history = model.history.history\nplt.title(\"loss on both train and test\")\nplt.plot(history[\"loss\"], label = \"train\")\nplt.plot(history[\"val_loss\"], label = \"test\")\nplt.legend()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"plt.title(\"accuracy on both train and test\")\nplt.plot(history[\"accuracy\"], label = \"train\")\nplt.plot(history[\"val_accuracy\"], label = \"test\")\nplt.legend()","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"result = model(x_test)\nx,y , _ = roc_curve(y_test, result)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# model(x_test[:200])\nresult = []\nfor i in range(501):\n    result.append(model(x_test[i:i+1]))\n","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"result = [i.numpy().flatten().item() for i in result]","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"x,y , _ = roc_curve(y_test, result)","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"\nplt.plot(x, y)\nplt.plot(x, x,color = \"orange\", ls = \":\")\nplt.title(\"auc = \" + str(roc_auc_score(y_test, result)))","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"confusion_matrix(y_test, np.round(result))","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(classification_report(y_test, np.round(result)))","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"we get 82% score on resnet and 78% on alexnet and 75% on lenet5","metadata":{}},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]}]}