state-of-art-models.ipynb

{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.7.12","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"import os\nfor dirname, _, filenames in os.walk('/kaggle/input'):\n    for filename in filenames:\n        print(os.path.join(dirname, filename))","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import pandas as pd \nimport numpy as np \nimport os\nimport cv2\nimport matplotlib.pyplot as plt\nimport warnings\n\nfrom tensorflow.keras.layers import Input, Lambda, Dense, Flatten, Dropout\nfrom tensorflow.keras.models import Model\nfrom tensorflow.keras.applications.vgg19 import preprocess_input\nfrom tensorflow.keras.preprocessing import image, image_dataset_from_directory\nfrom tensorflow.keras.preprocessing.image import ImageDataGenerator\nfrom tensorflow.keras.models import Sequential\nfrom tensorflow import keras\nimport tensorflow \n\nfrom tensorflow.keras.applications.vgg16 import VGG16 # VGG16\nfrom tensorflow.keras.applications.vgg19 import VGG19 # VGG19\nfrom tensorflow.keras.applications.resnet50 import ResNet50 # ResNet50\nfrom tensorflow.keras.applications import ResNet101 # ResNet 101\nfrom tensorflow.keras.applications.xception import Xception # Xception\nfrom tensorflow.keras.applications.mobilenet import MobileNet # MobileNet\nfrom tensorflow.keras.applications.densenet import DenseNet169 # DenseNet169\nfrom tensorflow.keras.applications.densenet import DenseNet121 # DenseNet121\nfrom tensorflow.keras.applications.mobilenet_v2 import MobileNetV2 # MobileNetV2\nfrom tensorflow.keras.applications.inception_v3 import InceptionV3 # InceptionV3\n\n\nimport scipy\nprint(\"Num GPUs Available: \", len(tensorflow.config.list_physical_devices('GPU')))\n\n\n# Set the seed value for experiment reproduci.bility.\nseed = 1842\ntensorflow.random.set_seed(seed)\nnp.random.seed(seed)\n# Turn off warnings for cleaner looking notebook\nwarnings.simplefilter('ignore')","metadata":{"execution":{"iopub.status.busy":"2022-08-06T13:32:12.248750Z","iopub.execute_input":"2022-08-06T13:32:12.249004Z","iopub.status.idle":"2022-08-06T13:32:18.882400Z","shell.execute_reply.started":"2022-08-06T13:32:12.248980Z","shell.execute_reply":"2022-08-06T13:32:18.881485Z"},"trusted":true},"execution_count":2,"outputs":[{"name":"stdout","text":"Num GPUs Available:  1\n","output_type":"stream"},{"name":"stderr","text":"2022-08-06 13:32:18.759991: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n2022-08-06 13:32:18.873387: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n2022-08-06 13:32:18.874142: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n","output_type":"stream"}]},{"cell_type":"code","source":"#define image dataset \n#why do we rescale?\nimage_generator = ImageDataGenerator(rescale=1/255, validation_split=0) #shear_range =.25, zoom_range =.2, horizontal_flip = True, rotation_range=20)     \n\n\ntrain_dataset = image_generator.flow_from_directory(batch_size=16,\n                                                 directory='/kaggle/input/alzheimers-dataset-4-class-of-images/Alzheimer_s Dataset/train',\n                                                 shuffle=True,\n                                                 target_size=(224,224), \n                                                 subset=\"training\",\n                                                 class_mode='categorical')\nimage_generator = ImageDataGenerator(rescale=1/255,validation_split=0.2) \nvalidation_dataset = image_generator.flow_from_directory(batch_size=16,\n                                                 directory='/kaggle/input/alzheimers-dataset-4-class-of-images/Alzheimer_s Dataset/test',\n                                                 shuffle=True,\n                                                 target_size=(224,224),\n                                                 class_mode='categorical')\n\n\nsubmission = image_generator.flow_from_directory(\n                                                 directory='/kaggle/input/alzheimers-dataset-4-class-of-images/Alzheimer_s Dataset/test',\n                                                 shuffle=False,\n                                                subset=\"validation\",\n                                                 target_size=(224,224), \n                                                 class_mode=None)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T13:32:18.884275Z","iopub.execute_input":"2022-08-06T13:32:18.884795Z","iopub.status.idle":"2022-08-06T13:32:24.722541Z","shell.execute_reply.started":"2022-08-06T13:32:18.884766Z","shell.execute_reply":"2022-08-06T13:32:24.721641Z"},"trusted":true},"execution_count":3,"outputs":[{"name":"stdout","text":"Found 5121 images belonging to 4 classes.\nFound 1279 images belonging to 4 classes.\nFound 254 images belonging to 4 classes.\n","output_type":"stream"}]},{"cell_type":"code","source":"#show flwoers for the first batch\n\nbatch_1_img = train_dataset[0]\nfor i in range(0,2):\n    img = batch_1_img[0][i]\n    lab = batch_1_img[1][i]\n    plt.imshow(img)\n    plt.title(lab)\n    plt.axis('off')\n    plt.show()","metadata":{"execution":{"iopub.status.busy":"2022-08-06T13:32:24.725224Z","iopub.execute_input":"2022-08-06T13:32:24.725653Z","iopub.status.idle":"2022-08-06T13:32:25.166439Z","shell.execute_reply.started":"2022-08-06T13:32:24.725615Z","shell.execute_reply":"2022-08-06T13:32:25.165488Z"},"trusted":true},"execution_count":4,"outputs":[{"output_type":"display_data","data":{"text/plain":"<Figure size 432x288 with 1 Axes>","image/png":"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\n"},"metadata":{"needs_background":"light"}},{"output_type":"display_data","data":{"text/plain":"<Figure size 432x288 with 1 Axes>","image/png":"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\n"},"metadata":{"needs_background":"light"}}]},{"cell_type":"markdown","source":"## VGG19","metadata":{}},{"cell_type":"code","source":"vgg = VGG19(input_shape=(224,224,3), weights='imagenet', include_top=False)\nfor layer in vgg.layers:\n    layer.trainable = False\nx = Flatten()(vgg.output)\n\nprediction = Dense(4, activation='softmax')(x)\n\nmodelvgg = Model(inputs=vgg.input, outputs=prediction)\nmodelvgg.summary()","metadata":{"execution":{"iopub.status.busy":"2022-08-06T13:32:25.169249Z","iopub.execute_input":"2022-08-06T13:32:25.169869Z","iopub.status.idle":"2022-08-06T13:32:29.741532Z","shell.execute_reply.started":"2022-08-06T13:32:25.169828Z","shell.execute_reply":"2022-08-06T13:32:29.740532Z"},"trusted":true},"execution_count":5,"outputs":[{"name":"stderr","text":"2022-08-06 13:32:25.214701: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 AVX512F FMA\nTo enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n2022-08-06 13:32:25.215270: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n2022-08-06 13:32:25.216751: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n2022-08-06 13:32:25.217765: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n2022-08-06 13:32:27.704335: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n2022-08-06 13:32:27.705237: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n2022-08-06 13:32:27.706058: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:937] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero\n2022-08-06 13:32:27.706784: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1510] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 15403 MB memory:  -> device: 0, name: Tesla P100-PCIE-16GB, pci bus id: 0000:00:04.0, compute capability: 6.0\n","output_type":"stream"},{"name":"stdout","text":"Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/vgg19/vgg19_weights_tf_dim_ordering_tf_kernels_notop.h5\n80142336/80134624 [==============================] - 1s 0us/step\n80150528/80134624 [==============================] - 1s 0us/step\nModel: \"model\"\n_________________________________________________________________\nLayer (type)                 Output Shape              Param #   \n=================================================================\ninput_1 (InputLayer)         [(None, 224, 224, 3)]     0         \n_________________________________________________________________\nblock1_conv1 (Conv2D)        (None, 224, 224, 64)      1792      \n_________________________________________________________________\nblock1_conv2 (Conv2D)        (None, 224, 224, 64)      36928     \n_________________________________________________________________\nblock1_pool (MaxPooling2D)   (None, 112, 112, 64)      0         \n_________________________________________________________________\nblock2_conv1 (Conv2D)        (None, 112, 112, 128)     73856     \n_________________________________________________________________\nblock2_conv2 (Conv2D)        (None, 112, 112, 128)     147584    \n_________________________________________________________________\nblock2_pool (MaxPooling2D)   (None, 56, 56, 128)       0         \n_________________________________________________________________\nblock3_conv1 (Conv2D)        (None, 56, 56, 256)       295168    \n_________________________________________________________________\nblock3_conv2 (Conv2D)        (None, 56, 56, 256)       590080    \n_________________________________________________________________\nblock3_conv3 (Conv2D)        (None, 56, 56, 256)       590080    \n_________________________________________________________________\nblock3_conv4 (Conv2D)        (None, 56, 56, 256)       590080    \n_________________________________________________________________\nblock3_pool (MaxPooling2D)   (None, 28, 28, 256)       0         \n_________________________________________________________________\nblock4_conv1 (Conv2D)        (None, 28, 28, 512)       1180160   \n_________________________________________________________________\nblock4_conv2 (Conv2D)        (None, 28, 28, 512)       2359808   \n_________________________________________________________________\nblock4_conv3 (Conv2D)        (None, 28, 28, 512)       2359808   \n_________________________________________________________________\nblock4_conv4 (Conv2D)        (None, 28, 28, 512)       2359808   \n_________________________________________________________________\nblock4_pool (MaxPooling2D)   (None, 14, 14, 512)       0         \n_________________________________________________________________\nblock5_conv1 (Conv2D)        (None, 14, 14, 512)       2359808   \n_________________________________________________________________\nblock5_conv2 (Conv2D)        (None, 14, 14, 512)       2359808   \n_________________________________________________________________\nblock5_conv3 (Conv2D)        (None, 14, 14, 512)       2359808   \n_________________________________________________________________\nblock5_conv4 (Conv2D)        (None, 14, 14, 512)       2359808   \n_________________________________________________________________\nblock5_pool (MaxPooling2D)   (None, 7, 7, 512)         0         \n_________________________________________________________________\nflatten (Flatten)            (None, 25088)             0         \n_________________________________________________________________\ndense (Dense)                (None, 4)                 100356    \n=================================================================\nTotal params: 20,124,740\nTrainable params: 100,356\nNon-trainable params: 20,024,384\n_________________________________________________________________\n","output_type":"stream"}]},{"cell_type":"code","source":"modelvgg.compile(optimizer='adam',\nloss=tensorflow.losses.CategoricalCrossentropy(),\nmetrics=[keras.metrics.AUC(name='auc'),'acc'])\ncallback = keras.callbacks.EarlyStopping(monitor='val_loss',\n                                            patience=8,\n                                            restore_best_weights=True)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T13:32:29.746059Z","iopub.execute_input":"2022-08-06T13:32:29.748076Z","iopub.status.idle":"2022-08-06T13:32:29.776529Z","shell.execute_reply.started":"2022-08-06T13:32:29.748037Z","shell.execute_reply":"2022-08-06T13:32:29.775516Z"},"trusted":true},"execution_count":6,"outputs":[]},{"cell_type":"code","source":"modelvgg.fit(train_dataset, epochs=50, validation_data=validation_dataset, callbacks=callback)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T13:32:29.777971Z","iopub.execute_input":"2022-08-06T13:32:29.778520Z","iopub.status.idle":"2022-08-06T13:36:20.477559Z","shell.execute_reply.started":"2022-08-06T13:32:29.778482Z","shell.execute_reply":"2022-08-06T13:36:20.476534Z"},"trusted":true},"execution_count":7,"outputs":[{"name":"stderr","text":"2022-08-06 13:32:29.907843: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:185] None of the MLIR Optimization Passes are enabled (registered 2)\n","output_type":"stream"},{"name":"stdout","text":"Epoch 1/50\n","output_type":"stream"},{"name":"stderr","text":"2022-08-06 13:32:32.313505: I tensorflow/stream_executor/cuda/cuda_dnn.cc:369] Loaded cuDNN version 8005\n","output_type":"stream"},{"name":"stdout","text":"321/321 [==============================] - 59s 154ms/step - loss: 0.9802 - auc: 0.8552 - acc: 0.6028 - val_loss: 0.7795 - val_auc: 0.8911 - val_acc: 0.6466\nEpoch 2/50\n321/321 [==============================] - 17s 53ms/step - loss: 0.6507 - auc: 0.9266 - acc: 0.7258 - val_loss: 0.8766 - val_auc: 0.8649 - val_acc: 0.6310\nEpoch 3/50\n321/321 [==============================] - 17s 54ms/step - loss: 0.4589 - auc: 0.9621 - acc: 0.8063 - val_loss: 0.7029 - val_auc: 0.9110 - val_acc: 0.7013\nEpoch 4/50\n321/321 [==============================] - 17s 53ms/step - loss: 0.3998 - auc: 0.9708 - acc: 0.8387 - val_loss: 0.8141 - val_auc: 0.8980 - val_acc: 0.6568\nEpoch 5/50\n321/321 [==============================] - 17s 53ms/step - loss: 0.3484 - auc: 0.9778 - acc: 0.8670 - val_loss: 1.7400 - val_auc: 0.8085 - val_acc: 0.4769\nEpoch 6/50\n321/321 [==============================] - 17s 53ms/step - loss: 0.2694 - auc: 0.9877 - acc: 0.9000 - val_loss: 0.9747 - val_auc: 0.8767 - val_acc: 0.6216\nEpoch 7/50\n321/321 [==============================] - 17s 53ms/step - loss: 0.2436 - auc: 0.9900 - acc: 0.9125 - val_loss: 0.8076 - val_auc: 0.9044 - val_acc: 0.6747\nEpoch 8/50\n321/321 [==============================] - 17s 53ms/step - loss: 0.2153 - auc: 0.9922 - acc: 0.9330 - val_loss: 1.0353 - val_auc: 0.8709 - val_acc: 0.6083\nEpoch 9/50\n321/321 [==============================] - 17s 53ms/step - loss: 0.1680 - auc: 0.9963 - acc: 0.9451 - val_loss: 0.7685 - val_auc: 0.9097 - val_acc: 0.6904\nEpoch 10/50\n321/321 [==============================] - 17s 53ms/step - loss: 0.1601 - auc: 0.9965 - acc: 0.9477 - val_loss: 0.8103 - val_auc: 0.9013 - val_acc: 0.6873\nEpoch 11/50\n321/321 [==============================] - 17s 53ms/step - loss: 0.1227 - auc: 0.9986 - acc: 0.9713 - val_loss: 0.8227 - val_auc: 0.9107 - val_acc: 0.6849\n","output_type":"stream"},{"execution_count":7,"output_type":"execute_result","data":{"text/plain":"<keras.callbacks.History at 0x7f3c70580690>"},"metadata":{}}]},{"cell_type":"code","source":"loss, auc,accuracy = modelvgg.evaluate(validation_dataset)\nprint(\"Loss: \", loss)\nprint(\"AUC: \", auc)\nprint(\"Accuracy: \", accuracy)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T13:36:20.478838Z","iopub.execute_input":"2022-08-06T13:36:20.479203Z","iopub.status.idle":"2022-08-06T13:36:24.111671Z","shell.execute_reply.started":"2022-08-06T13:36:20.479168Z","shell.execute_reply":"2022-08-06T13:36:24.110284Z"},"trusted":true},"execution_count":8,"outputs":[{"name":"stdout","text":"80/80 [==============================] - 4s 44ms/step - loss: 0.7029 - auc: 0.9110 - acc: 0.7013\nLoss:  0.7028596997261047\nAUC:  0.9110348224639893\nAccuracy:  0.7013291716575623\n","output_type":"stream"}]},{"cell_type":"markdown","source":"## VGG16","metadata":{}},{"cell_type":"code","source":"vgg = VGG16(input_shape=(224,224,3), weights='imagenet', include_top=False)\nfor layer in vgg.layers:\n    layer.trainable = False\nx = Flatten()(vgg.output)\n\nprediction = Dense(4, activation='softmax')(x)\n\nmodelvgg = Model(inputs=vgg.input, outputs=prediction)\nmodelvgg.summary()","metadata":{"execution":{"iopub.status.busy":"2022-08-06T13:36:24.112970Z","iopub.execute_input":"2022-08-06T13:36:24.113828Z","iopub.status.idle":"2022-08-06T13:36:24.820442Z","shell.execute_reply.started":"2022-08-06T13:36:24.113789Z","shell.execute_reply":"2022-08-06T13:36:24.819397Z"},"trusted":true},"execution_count":9,"outputs":[{"name":"stdout","text":"Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/vgg16/vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5\n58892288/58889256 [==============================] - 0s 0us/step\n58900480/58889256 [==============================] - 0s 0us/step\nModel: \"model_1\"\n_________________________________________________________________\nLayer (type)                 Output Shape              Param #   \n=================================================================\ninput_2 (InputLayer)         [(None, 224, 224, 3)]     0         \n_________________________________________________________________\nblock1_conv1 (Conv2D)        (None, 224, 224, 64)      1792      \n_________________________________________________________________\nblock1_conv2 (Conv2D)        (None, 224, 224, 64)      36928     \n_________________________________________________________________\nblock1_pool (MaxPooling2D)   (None, 112, 112, 64)      0         \n_________________________________________________________________\nblock2_conv1 (Conv2D)        (None, 112, 112, 128)     73856     \n_________________________________________________________________\nblock2_conv2 (Conv2D)        (None, 112, 112, 128)     147584    \n_________________________________________________________________\nblock2_pool (MaxPooling2D)   (None, 56, 56, 128)       0         \n_________________________________________________________________\nblock3_conv1 (Conv2D)        (None, 56, 56, 256)       295168    \n_________________________________________________________________\nblock3_conv2 (Conv2D)        (None, 56, 56, 256)       590080    \n_________________________________________________________________\nblock3_conv3 (Conv2D)        (None, 56, 56, 256)       590080    \n_________________________________________________________________\nblock3_pool (MaxPooling2D)   (None, 28, 28, 256)       0         \n_________________________________________________________________\nblock4_conv1 (Conv2D)        (None, 28, 28, 512)       1180160   \n_________________________________________________________________\nblock4_conv2 (Conv2D)        (None, 28, 28, 512)       2359808   \n_________________________________________________________________\nblock4_conv3 (Conv2D)        (None, 28, 28, 512)       2359808   \n_________________________________________________________________\nblock4_pool (MaxPooling2D)   (None, 14, 14, 512)       0         \n_________________________________________________________________\nblock5_conv1 (Conv2D)        (None, 14, 14, 512)       2359808   \n_________________________________________________________________\nblock5_conv2 (Conv2D)        (None, 14, 14, 512)       2359808   \n_________________________________________________________________\nblock5_conv3 (Conv2D)        (None, 14, 14, 512)       2359808   \n_________________________________________________________________\nblock5_pool (MaxPooling2D)   (None, 7, 7, 512)         0         \n_________________________________________________________________\nflatten_1 (Flatten)          (None, 25088)             0         \n_________________________________________________________________\ndense_1 (Dense)              (None, 4)                 100356    \n=================================================================\nTotal params: 14,815,044\nTrainable params: 100,356\nNon-trainable params: 14,714,688\n_________________________________________________________________\n","output_type":"stream"}]},{"cell_type":"code","source":"modelvgg.compile(optimizer='adam',\nloss=tensorflow.losses.CategoricalCrossentropy(),\nmetrics=[keras.metrics.AUC(name='auc'),'acc'])\ncallback = keras.callbacks.EarlyStopping(monitor='val_loss',\n                                            patience=8,\n                                            restore_best_weights=True)\n\nmodelvgg.fit(train_dataset, epochs=50, validation_data=validation_dataset, callbacks=callback)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T13:36:24.821819Z","iopub.execute_input":"2022-08-06T13:36:24.822428Z","iopub.status.idle":"2022-08-06T13:39:08.615706Z","shell.execute_reply.started":"2022-08-06T13:36:24.822390Z","shell.execute_reply":"2022-08-06T13:39:08.614647Z"},"trusted":true},"execution_count":10,"outputs":[{"name":"stdout","text":"Epoch 1/50\n321/321 [==============================] - 16s 46ms/step - loss: 0.8770 - auc: 0.8727 - acc: 0.6298 - val_loss: 0.7789 - val_auc: 0.8949 - val_acc: 0.6654\nEpoch 2/50\n321/321 [==============================] - 15s 47ms/step - loss: 0.5771 - auc: 0.9400 - acc: 0.7666 - val_loss: 0.8149 - val_auc: 0.8974 - val_acc: 0.6646\nEpoch 3/50\n321/321 [==============================] - 14s 45ms/step - loss: 0.3931 - auc: 0.9728 - acc: 0.8487 - val_loss: 0.7392 - val_auc: 0.9071 - val_acc: 0.6747\nEpoch 4/50\n321/321 [==============================] - 15s 46ms/step - loss: 0.2978 - auc: 0.9856 - acc: 0.8893 - val_loss: 0.8763 - val_auc: 0.8961 - val_acc: 0.6708\nEpoch 5/50\n321/321 [==============================] - 15s 46ms/step - loss: 0.2436 - auc: 0.9905 - acc: 0.9190 - val_loss: 0.8292 - val_auc: 0.9052 - val_acc: 0.6904\nEpoch 6/50\n321/321 [==============================] - 15s 46ms/step - loss: 0.1867 - auc: 0.9958 - acc: 0.9434 - val_loss: 0.8696 - val_auc: 0.9017 - val_acc: 0.6873\nEpoch 7/50\n321/321 [==============================] - 15s 45ms/step - loss: 0.1715 - auc: 0.9960 - acc: 0.9484 - val_loss: 0.9615 - val_auc: 0.8943 - val_acc: 0.6826\nEpoch 8/50\n321/321 [==============================] - 15s 47ms/step - loss: 0.1362 - auc: 0.9980 - acc: 0.9650 - val_loss: 1.0491 - val_auc: 0.8914 - val_acc: 0.6607\nEpoch 9/50\n321/321 [==============================] - 15s 46ms/step - loss: 0.1074 - auc: 0.9992 - acc: 0.9750 - val_loss: 0.9157 - val_auc: 0.9018 - val_acc: 0.6935\nEpoch 10/50\n321/321 [==============================] - 15s 46ms/step - loss: 0.0837 - auc: 0.9998 - acc: 0.9875 - val_loss: 0.9677 - val_auc: 0.9014 - val_acc: 0.6873\nEpoch 11/50\n321/321 [==============================] - 15s 47ms/step - loss: 0.1256 - auc: 0.9974 - acc: 0.9611 - val_loss: 1.0630 - val_auc: 0.8946 - val_acc: 0.6810\n","output_type":"stream"},{"execution_count":10,"output_type":"execute_result","data":{"text/plain":"<keras.callbacks.History at 0x7f3c10ffa190>"},"metadata":{}}]},{"cell_type":"code","source":"loss, auc, accuracy = modelvgg.evaluate(validation_dataset)\nprint(\"Loss: \", loss)\nprint(\"AUC: \", auc)\nprint(\"Accuracy: \", accuracy)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T13:39:08.620059Z","iopub.execute_input":"2022-08-06T13:39:08.620347Z","iopub.status.idle":"2022-08-06T13:39:13.797694Z","shell.execute_reply.started":"2022-08-06T13:39:08.620321Z","shell.execute_reply":"2022-08-06T13:39:13.796140Z"},"trusted":true},"execution_count":11,"outputs":[{"name":"stdout","text":"80/80 [==============================] - 3s 37ms/step - loss: 0.7392 - auc: 0.9071 - acc: 0.6747\nLoss:  0.7391741871833801\nAUC:  0.9070690274238586\nAccuracy:  0.6747459173202515\n","output_type":"stream"}]},{"cell_type":"markdown","source":"## ResNet50","metadata":{}},{"cell_type":"code","source":"rn = ResNet50(input_shape=(224,224,3), weights='imagenet', include_top=False)\nfor layer in rn.layers:\n    layer.trainable = False\nx = Flatten()(rn.output)\n\nprediction = Dense(4, activation='softmax')(x)\n\nmodelvgg = Model(inputs=rn.input, outputs=prediction)\n\nmodelvgg.compile(optimizer='adam',\nloss=tensorflow.losses.CategoricalCrossentropy(),\nmetrics=[keras.metrics.AUC(name='auc'),'acc'])\ncallback = keras.callbacks.EarlyStopping(monitor='val_loss',\n                                            patience=8,\n                                            restore_best_weights=True)\n\nmodelvgg.fit(train_dataset, epochs=50, validation_data=validation_dataset, callbacks=callback)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T14:50:54.086015Z","iopub.execute_input":"2022-08-06T14:50:54.086389Z","iopub.status.idle":"2022-08-06T14:55:49.106405Z","shell.execute_reply.started":"2022-08-06T14:50:54.086358Z","shell.execute_reply":"2022-08-06T14:55:49.105496Z"},"trusted":true},"execution_count":32,"outputs":[{"name":"stdout","text":"Epoch 1/50\n321/321 [==============================] - 20s 52ms/step - loss: 1.9128 - auc: 0.7576 - acc: 0.4829 - val_loss: 1.4363 - val_auc: 0.7513 - val_acc: 0.4996\nEpoch 2/50\n321/321 [==============================] - 16s 48ms/step - loss: 1.5107 - auc: 0.8061 - acc: 0.5462 - val_loss: 1.1503 - val_auc: 0.8233 - val_acc: 0.5723\nEpoch 3/50\n321/321 [==============================] - 16s 49ms/step - loss: 1.3908 - auc: 0.8248 - acc: 0.5751 - val_loss: 2.0616 - val_auc: 0.7541 - val_acc: 0.3690\nEpoch 4/50\n321/321 [==============================] - 16s 49ms/step - loss: 1.2988 - auc: 0.8423 - acc: 0.5860 - val_loss: 0.9961 - val_auc: 0.8439 - val_acc: 0.5817\nEpoch 5/50\n321/321 [==============================] - 15s 48ms/step - loss: 1.1138 - auc: 0.8648 - acc: 0.6214 - val_loss: 1.8263 - val_auc: 0.6969 - val_acc: 0.3948\nEpoch 6/50\n321/321 [==============================] - 15s 48ms/step - loss: 1.0702 - auc: 0.8710 - acc: 0.6501 - val_loss: 1.0534 - val_auc: 0.8511 - val_acc: 0.5903\nEpoch 7/50\n321/321 [==============================] - 16s 50ms/step - loss: 0.9988 - auc: 0.8847 - acc: 0.6626 - val_loss: 2.8191 - val_auc: 0.7065 - val_acc: 0.3675\nEpoch 8/50\n321/321 [==============================] - 16s 49ms/step - loss: 1.0490 - auc: 0.8802 - acc: 0.6577 - val_loss: 1.7045 - val_auc: 0.8078 - val_acc: 0.5051\nEpoch 9/50\n321/321 [==============================] - 16s 49ms/step - loss: 0.8173 - auc: 0.9058 - acc: 0.6936 - val_loss: 1.2309 - val_auc: 0.8399 - val_acc: 0.5950\nEpoch 10/50\n321/321 [==============================] - 16s 50ms/step - loss: 0.8489 - auc: 0.9080 - acc: 0.7049 - val_loss: 0.9953 - val_auc: 0.8609 - val_acc: 0.6200\nEpoch 11/50\n321/321 [==============================] - 16s 49ms/step - loss: 1.1082 - auc: 0.8825 - acc: 0.6733 - val_loss: 1.3654 - val_auc: 0.8302 - val_acc: 0.5442\nEpoch 12/50\n321/321 [==============================] - 16s 50ms/step - loss: 1.1932 - auc: 0.8768 - acc: 0.6583 - val_loss: 1.5912 - val_auc: 0.8050 - val_acc: 0.5426\nEpoch 13/50\n321/321 [==============================] - 16s 48ms/step - loss: 0.8462 - auc: 0.9170 - acc: 0.7135 - val_loss: 1.1617 - val_auc: 0.8305 - val_acc: 0.5715\nEpoch 14/50\n321/321 [==============================] - 16s 48ms/step - loss: 1.0175 - auc: 0.9022 - acc: 0.7006 - val_loss: 2.2822 - val_auc: 0.8213 - val_acc: 0.5559\nEpoch 15/50\n321/321 [==============================] - 16s 50ms/step - loss: 0.7028 - auc: 0.9407 - acc: 0.7575 - val_loss: 1.1603 - val_auc: 0.8584 - val_acc: 0.6294\nEpoch 16/50\n321/321 [==============================] - 16s 48ms/step - loss: 0.8321 - auc: 0.9193 - acc: 0.7278 - val_loss: 1.2709 - val_auc: 0.8104 - val_acc: 0.5434\nEpoch 17/50\n321/321 [==============================] - 16s 48ms/step - loss: 0.5983 - auc: 0.9442 - acc: 0.7770 - val_loss: 1.2105 - val_auc: 0.8363 - val_acc: 0.5754\nEpoch 18/50\n321/321 [==============================] - 16s 49ms/step - loss: 0.9182 - auc: 0.9116 - acc: 0.7182 - val_loss: 3.4421 - val_auc: 0.6496 - val_acc: 0.3706\n","output_type":"stream"},{"execution_count":32,"output_type":"execute_result","data":{"text/plain":"<keras.callbacks.History at 0x7f3774d02250>"},"metadata":{}}]},{"cell_type":"code","source":"loss, accuracy = modelvgg.evaluate(validation_dataset)\nprint(\"Loss: \", loss)\nprint(\"Accuracy: \", accuracy)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T13:41:59.713038Z","iopub.execute_input":"2022-08-06T13:41:59.714038Z","iopub.status.idle":"2022-08-06T13:42:02.994792Z","shell.execute_reply.started":"2022-08-06T13:41:59.714000Z","shell.execute_reply":"2022-08-06T13:42:02.993824Z"},"trusted":true},"execution_count":13,"outputs":[{"name":"stdout","text":"80/80 [==============================] - 3s 40ms/step - loss: 1.0272 - auc: 0.8430\nLoss:  1.0272291898727417\nAccuracy:  0.8429989218711853\n","output_type":"stream"}]},{"cell_type":"markdown","source":"## ResNet 101","metadata":{}},{"cell_type":"code","source":"rn = ResNet101(input_shape=(224,224,3), weights='imagenet', include_top=False)\nfor layer in rn.layers:\n    layer.trainable = False\nx = Flatten()(rn.output)\n\nprediction = Dense(4, activation='softmax')(x)\n\nmodelvgg = Model(inputs=rn.input, outputs=prediction)\n\nmodelvgg.compile(optimizer='adam',\nloss=tensorflow.losses.CategoricalCrossentropy(),\nmetrics=[keras.metrics.AUC(name='auc'),'acc'])\ncallback = keras.callbacks.EarlyStopping(monitor='val_loss',\n                                            patience=8,\n                                            restore_best_weights=True)\n\nmodelvgg.fit(train_dataset, epochs=50, validation_data=validation_dataset, callbacks=callback)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T13:51:24.930583Z","iopub.execute_input":"2022-08-06T13:51:24.931260Z","iopub.status.idle":"2022-08-06T14:01:25.830568Z","shell.execute_reply.started":"2022-08-06T13:51:24.931224Z","shell.execute_reply":"2022-08-06T14:01:25.829567Z"},"trusted":true},"execution_count":16,"outputs":[{"name":"stdout","text":"Epoch 1/50\n321/321 [==============================] - 32s 83ms/step - loss: 1.6128 - auc: 0.7657 - acc: 0.4814 - val_loss: 1.0784 - val_auc: 0.8224 - val_acc: 0.5246\nEpoch 2/50\n321/321 [==============================] - 24s 76ms/step - loss: 1.2686 - auc: 0.8232 - acc: 0.5602 - val_loss: 2.4120 - val_auc: 0.6374 - val_acc: 0.3698\nEpoch 3/50\n321/321 [==============================] - 25s 77ms/step - loss: 1.0984 - auc: 0.8481 - acc: 0.5919 - val_loss: 1.4691 - val_auc: 0.7832 - val_acc: 0.4300\nEpoch 4/50\n321/321 [==============================] - 24s 76ms/step - loss: 0.9649 - auc: 0.8706 - acc: 0.6210 - val_loss: 1.2721 - val_auc: 0.8391 - val_acc: 0.5379\nEpoch 5/50\n321/321 [==============================] - 24s 76ms/step - loss: 1.1699 - auc: 0.8593 - acc: 0.6159 - val_loss: 1.9416 - val_auc: 0.8096 - val_acc: 0.5927\nEpoch 6/50\n321/321 [==============================] - 25s 77ms/step - loss: 0.8559 - auc: 0.8964 - acc: 0.6825 - val_loss: 1.2061 - val_auc: 0.8302 - val_acc: 0.5567\nEpoch 7/50\n321/321 [==============================] - 24s 76ms/step - loss: 0.8781 - auc: 0.8995 - acc: 0.6881 - val_loss: 1.2443 - val_auc: 0.8356 - val_acc: 0.5919\nEpoch 8/50\n321/321 [==============================] - 25s 76ms/step - loss: 0.9084 - auc: 0.8953 - acc: 0.6813 - val_loss: 1.0700 - val_auc: 0.8300 - val_acc: 0.5762\nEpoch 9/50\n321/321 [==============================] - 25s 77ms/step - loss: 0.7574 - auc: 0.9134 - acc: 0.7006 - val_loss: 1.9648 - val_auc: 0.7843 - val_acc: 0.5379\nEpoch 10/50\n321/321 [==============================] - 24s 76ms/step - loss: 0.6060 - auc: 0.9399 - acc: 0.7596 - val_loss: 1.1118 - val_auc: 0.8082 - val_acc: 0.5340\nEpoch 11/50\n321/321 [==============================] - 25s 77ms/step - loss: 0.6322 - auc: 0.9346 - acc: 0.7458 - val_loss: 1.1703 - val_auc: 0.8438 - val_acc: 0.5559\nEpoch 12/50\n321/321 [==============================] - 25s 77ms/step - loss: 0.6227 - auc: 0.9378 - acc: 0.7489 - val_loss: 1.8479 - val_auc: 0.7594 - val_acc: 0.4363\nEpoch 13/50\n321/321 [==============================] - 24s 76ms/step - loss: 0.7356 - auc: 0.9261 - acc: 0.7405 - val_loss: 1.8618 - val_auc: 0.8253 - val_acc: 0.5856\nEpoch 14/50\n321/321 [==============================] - 24s 76ms/step - loss: 0.5532 - auc: 0.9506 - acc: 0.7875 - val_loss: 1.5266 - val_auc: 0.8205 - val_acc: 0.5684\nEpoch 15/50\n321/321 [==============================] - 24s 76ms/step - loss: 0.5787 - auc: 0.9543 - acc: 0.7987 - val_loss: 2.6650 - val_auc: 0.6790 - val_acc: 0.3823\nEpoch 16/50\n321/321 [==============================] - 25s 77ms/step - loss: 0.7634 - auc: 0.9315 - acc: 0.7618 - val_loss: 1.0697 - val_auc: 0.8580 - val_acc: 0.6020\nEpoch 17/50\n321/321 [==============================] - 25s 77ms/step - loss: 0.4850 - auc: 0.9583 - acc: 0.8043 - val_loss: 1.2639 - val_auc: 0.8320 - val_acc: 0.5551\nEpoch 18/50\n321/321 [==============================] - 24s 76ms/step - loss: 0.4372 - auc: 0.9656 - acc: 0.8182 - val_loss: 1.4653 - val_auc: 0.7961 - val_acc: 0.5199\nEpoch 19/50\n321/321 [==============================] - 25s 77ms/step - loss: 0.5085 - auc: 0.9584 - acc: 0.8198 - val_loss: 1.2275 - val_auc: 0.8476 - val_acc: 0.6028\nEpoch 20/50\n321/321 [==============================] - 24s 76ms/step - loss: 0.4182 - auc: 0.9675 - acc: 0.8350 - val_loss: 1.6978 - val_auc: 0.7952 - val_acc: 0.5059\nEpoch 21/50\n321/321 [==============================] - 25s 77ms/step - loss: 0.6097 - auc: 0.9535 - acc: 0.7952 - val_loss: 1.4445 - val_auc: 0.8408 - val_acc: 0.6099\nEpoch 22/50\n321/321 [==============================] - 24s 76ms/step - loss: 0.4202 - auc: 0.9681 - acc: 0.8342 - val_loss: 1.8460 - val_auc: 0.8243 - val_acc: 0.5958\nEpoch 23/50\n321/321 [==============================] - 25s 76ms/step - loss: 0.4416 - auc: 0.9662 - acc: 0.8323 - val_loss: 2.6322 - val_auc: 0.7471 - val_acc: 0.4832\nEpoch 24/50\n321/321 [==============================] - 25s 77ms/step - loss: 0.4914 - auc: 0.9626 - acc: 0.8237 - val_loss: 1.2737 - val_auc: 0.8145 - val_acc: 0.5676\n","output_type":"stream"},{"execution_count":16,"output_type":"execute_result","data":{"text/plain":"<keras.callbacks.History at 0x7f3be67e6a90>"},"metadata":{}}]},{"cell_type":"code","source":"loss, auc,accuracy = modelvgg.evaluate(validation_dataset)\nprint(\"Loss: \", loss)\nprint(\"AUC: \", auc)\nprint(\"Accuracy: \", accuracy)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T14:01:25.832432Z","iopub.execute_input":"2022-08-06T14:01:25.833309Z","iopub.status.idle":"2022-08-06T14:01:31.031298Z","shell.execute_reply.started":"2022-08-06T14:01:25.833269Z","shell.execute_reply":"2022-08-06T14:01:31.030275Z"},"trusted":true},"execution_count":17,"outputs":[{"name":"stdout","text":"80/80 [==============================] - 5s 64ms/step - loss: 1.0697 - auc: 0.8580 - acc: 0.6020\nLoss:  1.0696675777435303\nAUC:  0.8580460548400879\nAccuracy:  0.6020328402519226\n","output_type":"stream"}]},{"cell_type":"markdown","source":"## Xception","metadata":{}},{"cell_type":"code","source":"model = Xception(input_shape=(224,224,3), weights='imagenet', include_top=False)\nfor layer in model.layers:\n    layer.trainable = False\nx = Flatten()(model.output)\n\nprediction = Dense(4, activation='softmax')(x)\n\nmodelvgg = Model(inputs=model.input, outputs=prediction)\n\nmodelvgg.compile(optimizer='adam',\nloss=tensorflow.losses.CategoricalCrossentropy(),\nmetrics=[keras.metrics.AUC(name='auc'),'acc'])\ncallback = keras.callbacks.EarlyStopping(monitor='val_loss',\n                                            patience=8,\n                                            restore_best_weights=True)\n\nmodelvgg.fit(train_dataset, epochs=50, validation_data=validation_dataset, callbacks=callback)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T14:22:13.008679Z","iopub.execute_input":"2022-08-06T14:22:13.009453Z","iopub.status.idle":"2022-08-06T14:25:47.701654Z","shell.execute_reply.started":"2022-08-06T14:22:13.009417Z","shell.execute_reply":"2022-08-06T14:25:47.700537Z"},"trusted":true},"execution_count":20,"outputs":[{"name":"stdout","text":"Epoch 1/50\n321/321 [==============================] - 22s 62ms/step - loss: 2.8666 - auc: 0.7851 - acc: 0.5593 - val_loss: 1.7824 - val_auc: 0.8246 - val_acc: 0.5825\nEpoch 2/50\n321/321 [==============================] - 19s 58ms/step - loss: 1.7325 - auc: 0.8761 - acc: 0.6846 - val_loss: 3.4487 - val_auc: 0.7451 - val_acc: 0.5020\nEpoch 3/50\n321/321 [==============================] - 18s 57ms/step - loss: 1.3751 - auc: 0.8990 - acc: 0.7391 - val_loss: 2.1722 - val_auc: 0.8262 - val_acc: 0.5942\nEpoch 4/50\n321/321 [==============================] - 19s 58ms/step - loss: 0.7832 - auc: 0.9445 - acc: 0.8110 - val_loss: 5.6367 - val_auc: 0.7241 - val_acc: 0.5410\nEpoch 5/50\n321/321 [==============================] - 18s 57ms/step - loss: 0.4337 - auc: 0.9739 - acc: 0.8778 - val_loss: 3.9394 - val_auc: 0.7867 - val_acc: 0.5848\nEpoch 6/50\n321/321 [==============================] - 18s 57ms/step - loss: 0.6458 - auc: 0.9576 - acc: 0.8453 - val_loss: 2.1337 - val_auc: 0.8291 - val_acc: 0.6083\nEpoch 7/50\n321/321 [==============================] - 19s 58ms/step - loss: 0.8141 - auc: 0.9560 - acc: 0.8506 - val_loss: 3.1173 - val_auc: 0.8059 - val_acc: 0.6005\nEpoch 8/50\n321/321 [==============================] - 18s 57ms/step - loss: 0.3721 - auc: 0.9816 - acc: 0.9117 - val_loss: 2.7496 - val_auc: 0.8227 - val_acc: 0.6200\nEpoch 9/50\n321/321 [==============================] - 18s 56ms/step - loss: 0.5044 - auc: 0.9724 - acc: 0.9020 - val_loss: 3.3935 - val_auc: 0.8116 - val_acc: 0.6005\n","output_type":"stream"},{"execution_count":20,"output_type":"execute_result","data":{"text/plain":"<keras.callbacks.History at 0x7f37941aaed0>"},"metadata":{}}]},{"cell_type":"code","source":"loss, auc,accuracy = modelvgg.evaluate(validation_dataset)\nprint(\"Loss: \", loss)\nprint(\"AUC: \", auc)\nprint(\"Accuracy: \", accuracy)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T14:34:35.020766Z","iopub.execute_input":"2022-08-06T14:34:35.021179Z","iopub.status.idle":"2022-08-06T14:34:38.889267Z","shell.execute_reply.started":"2022-08-06T14:34:35.021145Z","shell.execute_reply":"2022-08-06T14:34:38.888296Z"},"trusted":true},"execution_count":21,"outputs":[{"name":"stdout","text":"80/80 [==============================] - 4s 47ms/step - loss: 1.7824 - auc: 0.8246 - acc: 0.5825\nLoss:  1.7824040651321411\nAUC:  0.8246198296546936\nAccuracy:  0.5824863314628601\n","output_type":"stream"}]},{"cell_type":"markdown","source":"## MobileNet","metadata":{}},{"cell_type":"code","source":"model = MobileNet(input_shape=(224,224,3), weights='imagenet', include_top=False)\nfor layer in model.layers:\n    layer.trainable = False\nx = Flatten()(model.output)\n\nprediction = Dense(4, activation='softmax')(x)\n\nmodelvgg = Model(inputs=model.input, outputs=prediction)\n\nmodelvgg.compile(optimizer='adam',\nloss=tensorflow.losses.CategoricalCrossentropy(),\nmetrics=[keras.metrics.AUC(name='auc'),'acc'])\ncallback = keras.callbacks.EarlyStopping(monitor='val_loss',\n                                            patience=8,\n                                            restore_best_weights=True)\n\nmodelvgg.fit(train_dataset, epochs=50, validation_data=validation_dataset, callbacks=callback)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T14:34:38.892678Z","iopub.execute_input":"2022-08-06T14:34:38.892980Z","iopub.status.idle":"2022-08-06T14:36:45.334783Z","shell.execute_reply.started":"2022-08-06T14:34:38.892953Z","shell.execute_reply":"2022-08-06T14:36:45.333616Z"},"trusted":true},"execution_count":22,"outputs":[{"name":"stdout","text":"Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/mobilenet/mobilenet_1_0_224_tf_no_top.h5\n17227776/17225924 [==============================] - 0s 0us/step\n17235968/17225924 [==============================] - 0s 0us/step\nEpoch 1/50\n321/321 [==============================] - 14s 39ms/step - loss: 2.5882 - auc: 0.8475 - acc: 0.6702 - val_loss: 2.6967 - val_auc: 0.8380 - val_acc: 0.6622\nEpoch 2/50\n321/321 [==============================] - 12s 37ms/step - loss: 0.6357 - auc: 0.9622 - acc: 0.8824 - val_loss: 2.4328 - val_auc: 0.8496 - val_acc: 0.6849\nEpoch 3/50\n321/321 [==============================] - 12s 39ms/step - loss: 0.3809 - auc: 0.9811 - acc: 0.9236 - val_loss: 3.3881 - val_auc: 0.8363 - val_acc: 0.6693\nEpoch 4/50\n321/321 [==============================] - 12s 37ms/step - loss: 0.2376 - auc: 0.9889 - acc: 0.9498 - val_loss: 3.1374 - val_auc: 0.8320 - val_acc: 0.6654\nEpoch 5/50\n321/321 [==============================] - 12s 37ms/step - loss: 0.2890 - auc: 0.9856 - acc: 0.9477 - val_loss: 6.0314 - val_auc: 0.7677 - val_acc: 0.5989\nEpoch 6/50\n321/321 [==============================] - 13s 39ms/step - loss: 0.1757 - auc: 0.9915 - acc: 0.9666 - val_loss: 4.7786 - val_auc: 0.8082 - val_acc: 0.6505\nEpoch 7/50\n321/321 [==============================] - 12s 38ms/step - loss: 0.1076 - auc: 0.9947 - acc: 0.9768 - val_loss: 3.9704 - val_auc: 0.8317 - val_acc: 0.6912\nEpoch 8/50\n321/321 [==============================] - 13s 41ms/step - loss: 0.2223 - auc: 0.9882 - acc: 0.9637 - val_loss: 4.2736 - val_auc: 0.8233 - val_acc: 0.6763\nEpoch 9/50\n321/321 [==============================] - 12s 39ms/step - loss: 0.4427 - auc: 0.9802 - acc: 0.9531 - val_loss: 10.4343 - val_auc: 0.7693 - val_acc: 0.6185\nEpoch 10/50\n321/321 [==============================] - 12s 39ms/step - loss: 0.5031 - auc: 0.9807 - acc: 0.9471 - val_loss: 5.8331 - val_auc: 0.8163 - val_acc: 0.6787\n","output_type":"stream"},{"execution_count":22,"output_type":"execute_result","data":{"text/plain":"<keras.callbacks.History at 0x7f377fec9c10>"},"metadata":{}}]},{"cell_type":"code","source":"loss, auc,accuracy = modelvgg.evaluate(validation_dataset)\nprint(\"Loss: \", loss)\nprint(\"AUC: \", auc)\nprint(\"Accuracy: \", accuracy)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T14:36:45.336491Z","iopub.execute_input":"2022-08-06T14:36:45.336862Z","iopub.status.idle":"2022-08-06T14:36:47.791278Z","shell.execute_reply.started":"2022-08-06T14:36:45.336826Z","shell.execute_reply":"2022-08-06T14:36:47.790275Z"},"trusted":true},"execution_count":23,"outputs":[{"name":"stdout","text":"80/80 [==============================] - 2s 29ms/step - loss: 2.4328 - auc: 0.8496 - acc: 0.6849\nLoss:  2.4328389167785645\nAUC:  0.8495726585388184\nAccuracy:  0.6849100589752197\n","output_type":"stream"}]},{"cell_type":"markdown","source":"## MobileNetV2","metadata":{}},{"cell_type":"code","source":"model = MobileNetV2(input_shape=(224,224,3), weights='imagenet', include_top=False)\nfor layer in model.layers:\n    layer.trainable = False\nx = Flatten()(model.output)\n\nprediction = Dense(4, activation='softmax')(x)\n\nmodelvgg = Model(inputs=model.input, outputs=prediction)\n\nmodelvgg.compile(optimizer='adam',\nloss=tensorflow.losses.CategoricalCrossentropy(),\nmetrics=[keras.metrics.AUC(name='auc'),'acc'])\ncallback = keras.callbacks.EarlyStopping(monitor='val_loss',\n                                            patience=8,\n                                            restore_best_weights=True)\n\nmodelvgg.fit(train_dataset, epochs=50, validation_data=validation_dataset, callbacks=callback)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T14:36:47.794569Z","iopub.execute_input":"2022-08-06T14:36:47.795511Z","iopub.status.idle":"2022-08-06T14:38:46.277532Z","shell.execute_reply.started":"2022-08-06T14:36:47.795472Z","shell.execute_reply":"2022-08-06T14:38:46.276610Z"},"trusted":true},"execution_count":24,"outputs":[{"name":"stdout","text":"Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/mobilenet_v2/mobilenet_v2_weights_tf_dim_ordering_tf_kernels_1.0_224_no_top.h5\n9412608/9406464 [==============================] - 0s 0us/step\n9420800/9406464 [==============================] - 0s 0us/step\nEpoch 1/50\n321/321 [==============================] - 17s 44ms/step - loss: 2.8531 - auc: 0.8252 - acc: 0.6255 - val_loss: 2.3895 - val_auc: 0.8196 - val_acc: 0.6138\nEpoch 2/50\n321/321 [==============================] - 13s 40ms/step - loss: 0.7955 - auc: 0.9494 - acc: 0.8391 - val_loss: 3.1490 - val_auc: 0.8184 - val_acc: 0.6419\nEpoch 3/50\n321/321 [==============================] - 12s 39ms/step - loss: 0.6076 - auc: 0.9642 - acc: 0.8858 - val_loss: 3.9881 - val_auc: 0.8101 - val_acc: 0.6497\nEpoch 4/50\n321/321 [==============================] - 12s 39ms/step - loss: 0.3596 - auc: 0.9828 - acc: 0.9264 - val_loss: 4.4053 - val_auc: 0.8115 - val_acc: 0.6396\nEpoch 5/50\n321/321 [==============================] - 13s 40ms/step - loss: 0.3044 - auc: 0.9843 - acc: 0.9369 - val_loss: 4.0007 - val_auc: 0.8065 - val_acc: 0.6443\nEpoch 6/50\n321/321 [==============================] - 12s 38ms/step - loss: 0.4331 - auc: 0.9776 - acc: 0.9194 - val_loss: 5.5549 - val_auc: 0.7983 - val_acc: 0.6474\nEpoch 7/50\n321/321 [==============================] - 12s 39ms/step - loss: 0.2927 - auc: 0.9851 - acc: 0.9428 - val_loss: 7.4929 - val_auc: 0.7875 - val_acc: 0.6372\nEpoch 8/50\n321/321 [==============================] - 13s 40ms/step - loss: 0.2954 - auc: 0.9848 - acc: 0.9506 - val_loss: 8.5050 - val_auc: 0.7989 - val_acc: 0.6599\nEpoch 9/50\n321/321 [==============================] - 12s 39ms/step - loss: 0.2123 - auc: 0.9889 - acc: 0.9652 - val_loss: 5.9850 - val_auc: 0.8022 - val_acc: 0.6560\n","output_type":"stream"},{"execution_count":24,"output_type":"execute_result","data":{"text/plain":"<keras.callbacks.History at 0x7f37f3277690>"},"metadata":{}}]},{"cell_type":"code","source":"loss, auc,accuracy = modelvgg.evaluate(validation_dataset)\nprint(\"Loss: \", loss)\nprint(\"AUC: \", auc)\nprint(\"Accuracy: \", accuracy)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T14:38:46.279111Z","iopub.execute_input":"2022-08-06T14:38:46.279468Z","iopub.status.idle":"2022-08-06T14:38:48.795294Z","shell.execute_reply.started":"2022-08-06T14:38:46.279433Z","shell.execute_reply":"2022-08-06T14:38:48.794209Z"},"trusted":true},"execution_count":25,"outputs":[{"name":"stdout","text":"80/80 [==============================] - 2s 30ms/step - loss: 2.3895 - auc: 0.8196 - acc: 0.6138\nLoss:  2.389451742172241\nAUC:  0.8196309208869934\nAccuracy:  0.613760769367218\n","output_type":"stream"}]},{"cell_type":"markdown","source":"## DenseNet169","metadata":{}},{"cell_type":"code","source":"model = DenseNet169(input_shape=(224,224,3), weights='imagenet', include_top=False)\nfor layer in model.layers:\n    layer.trainable = False\nx = Flatten()(model.output)\n\nprediction = Dense(4, activation='softmax')(x)\n\nmodelvgg = Model(inputs=model.input, outputs=prediction)\n\nmodelvgg.compile(optimizer='adam',\nloss=tensorflow.losses.CategoricalCrossentropy(),\nmetrics=[keras.metrics.AUC(name='auc'),'acc'])\ncallback = keras.callbacks.EarlyStopping(monitor='val_loss',\n                                            patience=8,\n                                            restore_best_weights=True)\n\nmodelvgg.fit(train_dataset, epochs=50, validation_data=validation_dataset, callbacks=callback)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T14:38:48.797008Z","iopub.execute_input":"2022-08-06T14:38:48.797771Z","iopub.status.idle":"2022-08-06T14:42:38.406213Z","shell.execute_reply.started":"2022-08-06T14:38:48.797732Z","shell.execute_reply":"2022-08-06T14:42:38.405308Z"},"trusted":true},"execution_count":26,"outputs":[{"name":"stdout","text":"Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/densenet/densenet169_weights_tf_dim_ordering_tf_kernels_notop.h5\n51879936/51877672 [==============================] - 0s 0us/step\n51888128/51877672 [==============================] - 0s 0us/step\nEpoch 1/50\n321/321 [==============================] - 30s 71ms/step - loss: 3.5551 - auc: 0.8000 - acc: 0.6075 - val_loss: 2.8930 - val_auc: 0.8131 - val_acc: 0.6153\nEpoch 2/50\n321/321 [==============================] - 19s 60ms/step - loss: 1.0477 - auc: 0.9295 - acc: 0.7922 - val_loss: 3.3776 - val_auc: 0.8068 - val_acc: 0.6208\nEpoch 3/50\n321/321 [==============================] - 19s 60ms/step - loss: 0.8506 - auc: 0.9461 - acc: 0.8332 - val_loss: 2.4722 - val_auc: 0.8282 - val_acc: 0.6302\nEpoch 4/50\n321/321 [==============================] - 20s 61ms/step - loss: 0.7997 - auc: 0.9518 - acc: 0.8541 - val_loss: 3.3638 - val_auc: 0.7947 - val_acc: 0.6083\nEpoch 5/50\n321/321 [==============================] - 19s 59ms/step - loss: 2.1009 - auc: 0.9092 - acc: 0.8080 - val_loss: 4.6307 - val_auc: 0.8086 - val_acc: 0.6380\nEpoch 6/50\n321/321 [==============================] - 20s 61ms/step - loss: 1.6248 - auc: 0.9414 - acc: 0.8639 - val_loss: 5.1417 - val_auc: 0.7892 - val_acc: 0.6325\nEpoch 7/50\n321/321 [==============================] - 19s 60ms/step - loss: 0.3257 - auc: 0.9828 - acc: 0.9365 - val_loss: 3.4849 - val_auc: 0.8271 - val_acc: 0.6693\nEpoch 8/50\n321/321 [==============================] - 19s 59ms/step - loss: 0.3687 - auc: 0.9813 - acc: 0.9336 - val_loss: 4.5859 - val_auc: 0.8155 - val_acc: 0.6661\nEpoch 9/50\n321/321 [==============================] - 20s 61ms/step - loss: 0.4679 - auc: 0.9782 - acc: 0.9307 - val_loss: 3.5752 - val_auc: 0.8274 - val_acc: 0.6599\nEpoch 10/50\n321/321 [==============================] - 19s 61ms/step - loss: 0.4814 - auc: 0.9754 - acc: 0.9254 - val_loss: 6.6079 - val_auc: 0.7979 - val_acc: 0.6466\nEpoch 11/50\n321/321 [==============================] - 19s 60ms/step - loss: 0.4162 - auc: 0.9782 - acc: 0.9385 - val_loss: 4.5710 - val_auc: 0.8168 - val_acc: 0.6607\n","output_type":"stream"},{"execution_count":26,"output_type":"execute_result","data":{"text/plain":"<keras.callbacks.History at 0x7f377ca35890>"},"metadata":{}}]},{"cell_type":"code","source":"loss, auc,accuracy = modelvgg.evaluate(validation_dataset)\nprint(\"Loss: \", loss)\nprint(\"AUC: \", auc)\nprint(\"Accuracy: \", accuracy)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T14:42:38.407492Z","iopub.execute_input":"2022-08-06T14:42:38.407887Z","iopub.status.idle":"2022-08-06T14:42:42.522843Z","shell.execute_reply.started":"2022-08-06T14:42:38.407850Z","shell.execute_reply":"2022-08-06T14:42:42.521935Z"},"trusted":true},"execution_count":27,"outputs":[{"name":"stdout","text":"80/80 [==============================] - 4s 50ms/step - loss: 2.4722 - auc: 0.8282 - acc: 0.6302\nLoss:  2.4722206592559814\nAUC:  0.8282344937324524\nAccuracy:  0.6301798224449158\n","output_type":"stream"}]},{"cell_type":"markdown","source":"## DenseNet121","metadata":{}},{"cell_type":"code","source":"model = DenseNet121(input_shape=(224,224,3), weights='imagenet', include_top=False)\nfor layer in model.layers:\n    layer.trainable = False\nx = Flatten()(model.output)\n\nprediction = Dense(4, activation='softmax')(x)\n\nmodelvgg = Model(inputs=model.input, outputs=prediction)\n\nmodelvgg.compile(optimizer='adam',\nloss=tensorflow.losses.CategoricalCrossentropy(),\nmetrics=[keras.metrics.AUC(name='auc'),'acc'])\ncallback = keras.callbacks.EarlyStopping(monitor='val_loss',\n                                            patience=8,\n                                            restore_best_weights=True)\n\nmodelvgg.fit(train_dataset, epochs=50, validation_data=validation_dataset, callbacks=callback)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T14:42:42.524792Z","iopub.execute_input":"2022-08-06T14:42:42.525400Z","iopub.status.idle":"2022-08-06T14:45:18.766558Z","shell.execute_reply.started":"2022-08-06T14:42:42.525362Z","shell.execute_reply":"2022-08-06T14:45:18.765555Z"},"trusted":true},"execution_count":28,"outputs":[{"name":"stdout","text":"Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/densenet/densenet121_weights_tf_dim_ordering_tf_kernels_notop.h5\n29089792/29084464 [==============================] - 0s 0us/step\n29097984/29084464 [==============================] - 0s 0us/step\nEpoch 1/50\n321/321 [==============================] - 24s 58ms/step - loss: 3.1259 - auc: 0.8040 - acc: 0.5870 - val_loss: 1.5951 - val_auc: 0.8411 - val_acc: 0.5872\nEpoch 2/50\n321/321 [==============================] - 15s 48ms/step - loss: 1.1252 - auc: 0.9146 - acc: 0.7534 - val_loss: 2.4084 - val_auc: 0.8276 - val_acc: 0.6130\nEpoch 3/50\n321/321 [==============================] - 16s 51ms/step - loss: 0.8534 - auc: 0.9406 - acc: 0.8110 - val_loss: 3.1793 - val_auc: 0.7636 - val_acc: 0.5207\nEpoch 4/50\n321/321 [==============================] - 16s 49ms/step - loss: 0.6591 - auc: 0.9581 - acc: 0.8489 - val_loss: 2.3553 - val_auc: 0.8272 - val_acc: 0.6185\nEpoch 5/50\n321/321 [==============================] - 16s 49ms/step - loss: 0.4431 - auc: 0.9751 - acc: 0.8869 - val_loss: 1.8995 - val_auc: 0.8470 - val_acc: 0.6458\nEpoch 6/50\n321/321 [==============================] - 16s 50ms/step - loss: 0.5802 - auc: 0.9678 - acc: 0.8906 - val_loss: 4.0385 - val_auc: 0.7936 - val_acc: 0.6052\nEpoch 7/50\n321/321 [==============================] - 16s 50ms/step - loss: 0.5367 - auc: 0.9693 - acc: 0.8977 - val_loss: 7.2852 - val_auc: 0.6562 - val_acc: 0.4300\nEpoch 8/50\n321/321 [==============================] - 17s 52ms/step - loss: 0.5565 - auc: 0.9686 - acc: 0.8973 - val_loss: 2.9011 - val_auc: 0.8274 - val_acc: 0.6575\nEpoch 9/50\n321/321 [==============================] - 16s 51ms/step - loss: 0.2020 - auc: 0.9907 - acc: 0.9479 - val_loss: 3.6287 - val_auc: 0.8048 - val_acc: 0.6286\n","output_type":"stream"},{"execution_count":28,"output_type":"execute_result","data":{"text/plain":"<keras.callbacks.History at 0x7f377cb756d0>"},"metadata":{}}]},{"cell_type":"code","source":"loss, auc,accuracy = modelvgg.evaluate(validation_dataset)\nprint(\"Loss: \", loss)\nprint(\"AUC: \", auc)\nprint(\"Accuracy: \", accuracy)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T14:45:18.767919Z","iopub.execute_input":"2022-08-06T14:45:18.768520Z","iopub.status.idle":"2022-08-06T14:45:23.962381Z","shell.execute_reply.started":"2022-08-06T14:45:18.768481Z","shell.execute_reply":"2022-08-06T14:45:23.961309Z"},"trusted":true},"execution_count":29,"outputs":[{"name":"stdout","text":"80/80 [==============================] - 3s 41ms/step - loss: 1.5951 - auc: 0.8411 - acc: 0.5872\nLoss:  1.59514319896698\nAUC:  0.8410608768463135\nAccuracy:  0.5871774554252625\n","output_type":"stream"}]},{"cell_type":"markdown","source":"## InceptionV3","metadata":{}},{"cell_type":"code","source":"model = InceptionV3(input_shape=(224,224,3), weights='imagenet', include_top=False)\nfor layer in model.layers:\n    layer.trainable = False\nx = Flatten()(model.output)\n\nprediction = Dense(4, activation='softmax')(x)\n\nmodelvgg = Model(inputs=model.input, outputs=prediction)\n\nmodelvgg.compile(optimizer='adam',\nloss=tensorflow.losses.CategoricalCrossentropy(),\nmetrics=[keras.metrics.AUC(name='auc'),'acc'])\ncallback = keras.callbacks.EarlyStopping(monitor='val_loss',\n                                            patience=8,\n                                            restore_best_weights=True)\n\nmodelvgg.fit(train_dataset, epochs=50, validation_data=validation_dataset, callbacks=callback)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T14:45:23.965635Z","iopub.execute_input":"2022-08-06T14:45:23.965972Z","iopub.status.idle":"2022-08-06T14:48:13.920714Z","shell.execute_reply.started":"2022-08-06T14:45:23.965945Z","shell.execute_reply":"2022-08-06T14:48:13.919529Z"},"trusted":true},"execution_count":30,"outputs":[{"name":"stdout","text":"Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/inception_v3/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5\n87916544/87910968 [==============================] - 0s 0us/step\n87924736/87910968 [==============================] - 0s 0us/step\nEpoch 1/50\n321/321 [==============================] - 21s 52ms/step - loss: 3.1882 - auc: 0.7908 - acc: 0.5786 - val_loss: 5.5198 - val_auc: 0.6398 - val_acc: 0.4066\nEpoch 2/50\n321/321 [==============================] - 15s 45ms/step - loss: 1.5306 - auc: 0.8909 - acc: 0.7284 - val_loss: 2.9862 - val_auc: 0.7425 - val_acc: 0.5043\nEpoch 3/50\n321/321 [==============================] - 14s 44ms/step - loss: 0.9635 - auc: 0.9341 - acc: 0.8047 - val_loss: 2.9028 - val_auc: 0.7588 - val_acc: 0.5184\nEpoch 4/50\n321/321 [==============================] - 15s 46ms/step - loss: 0.6687 - auc: 0.9599 - acc: 0.8606 - val_loss: 3.4409 - val_auc: 0.7641 - val_acc: 0.5309\nEpoch 5/50\n321/321 [==============================] - 15s 45ms/step - loss: 1.2785 - auc: 0.9319 - acc: 0.8246 - val_loss: 4.1383 - val_auc: 0.7788 - val_acc: 0.5700\nEpoch 6/50\n321/321 [==============================] - 14s 44ms/step - loss: 0.6337 - auc: 0.9623 - acc: 0.8748 - val_loss: 3.1259 - val_auc: 0.8027 - val_acc: 0.5989\nEpoch 7/50\n321/321 [==============================] - 15s 47ms/step - loss: 0.2448 - auc: 0.9881 - acc: 0.9385 - val_loss: 4.8431 - val_auc: 0.7594 - val_acc: 0.5598\nEpoch 8/50\n321/321 [==============================] - 14s 45ms/step - loss: 0.6966 - auc: 0.9639 - acc: 0.8928 - val_loss: 5.6922 - val_auc: 0.7623 - val_acc: 0.5872\nEpoch 9/50\n321/321 [==============================] - 15s 46ms/step - loss: 0.2994 - auc: 0.9857 - acc: 0.9375 - val_loss: 5.7895 - val_auc: 0.7537 - val_acc: 0.5723\nEpoch 10/50\n321/321 [==============================] - 14s 44ms/step - loss: 0.3180 - auc: 0.9842 - acc: 0.9395 - val_loss: 6.8114 - val_auc: 0.7358 - val_acc: 0.5567\nEpoch 11/50\n321/321 [==============================] - 15s 47ms/step - loss: 0.4206 - auc: 0.9772 - acc: 0.9211 - val_loss: 5.4713 - val_auc: 0.7623 - val_acc: 0.5653\n","output_type":"stream"},{"execution_count":30,"output_type":"execute_result","data":{"text/plain":"<keras.callbacks.History at 0x7f377c975190>"},"metadata":{}}]},{"cell_type":"code","source":"loss, auc,accuracy = modelvgg.evaluate(validation_dataset)\nprint(\"Loss: \", loss)\nprint(\"AUC: \", auc)\nprint(\"Accuracy: \", accuracy)","metadata":{"execution":{"iopub.status.busy":"2022-08-06T14:48:13.922398Z","iopub.execute_input":"2022-08-06T14:48:13.922874Z","iopub.status.idle":"2022-08-06T14:48:16.917581Z","shell.execute_reply.started":"2022-08-06T14:48:13.922832Z","shell.execute_reply":"2022-08-06T14:48:16.916670Z"},"trusted":true},"execution_count":31,"outputs":[{"name":"stdout","text":"80/80 [==============================] - 3s 36ms/step - loss: 2.9028 - auc: 0.7588 - acc: 0.5184\nLoss:  2.90277099609375\nAUC:  0.7588043212890625\nAccuracy:  0.5183737277984619\n","output_type":"stream"}]}]}