LargerDataset_Covid19_Pred.ipynb

{"cells":[{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":23191,"status":"ok","timestamp":1665682231989,"user":{"displayName":"Akash Das","userId":"13248720121128963302"},"user_tz":-330},"id":"r_QCwymVq3tn","outputId":"1b3677f5-e50c-4002-ef3a-53066d89f52f"},"outputs":[{"output_type":"stream","name":"stdout","text":["Mounted at /content/drive\n"]}],"source":["from google.colab import drive\n","drive.mount('/content/drive')"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"N3cA-9aVrDbq"},"outputs":[],"source":["import numpy as np\n","import tensorflow as tf\n","import cv2\n","from sklearn.model_selection import train_test_split\n","from tensorflow import keras\n","from sklearn.metrics import roc_curve\n","from sklearn.metrics import roc_auc_score\n","import seaborn as sns\n","from sklearn.metrics import classification_report,confusion_matrix"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"LjtiAyVsxzFn"},"outputs":[],"source":["y = np.zeros(600)\n","y[:300] = 0\n","y[300:600] = 1 # 1 = +ve"]},{"cell_type":"code","source":["#base 64 to image cvt"],"metadata":{"id":"XjZAO1t7VgPH"},"execution_count":null,"outputs":[]},{"cell_type":"code","execution_count":null,"metadata":{"id":"oEpivlEeyI13"},"outputs":[],"source":["X = np.zeros((600,256,256,3))\n","\n","#normal first\n","\n","for i in range(300):\n","  img = cv2.imread(f\"/content/drive/MyDrive/Covid19/Newron Covid 19/Normal_Large/Normal-{i+1}.png\") #cvt to greyscale by 0....Ada path\n","  img = cv2.resize(img, (256,256))\n","  X[i] = img.copy()\n","\n","for i in range(300):\n","  img = cv2.imread(f\"/content/drive/MyDrive/Covid19/Newron Covid 19/Covid_Large/COVID-{i+1}.png\") #cvt to greyscale by 0....Ada path\n","  img = cv2.resize(img, (256,256))\n","  X[i+100] = img.copy() # storing after 100, first 100= normal CXR\n","\n","X = X/255.0"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"kgvfI8rq023M"},"outputs":[],"source":["X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"QBzBCx8P4OXP"},"outputs":[],"source":["from tensorflow import keras\n","from keras import layers\n","\n","data_augmentation = keras.Sequential(\n","    [layers.RandomFlip(\"horizontal\"), layers.RandomRotation(0.1),]\n",")"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":3690,"status":"ok","timestamp":1665682742992,"user":{"displayName":"Akash Das","userId":"13248720121128963302"},"user_tz":-330},"id":"_y-nVs4F3fLC","outputId":"e5a812fe-7c5b-4b75-8fd0-0ae5bd89b6b6"},"outputs":[{"output_type":"stream","name":"stdout","text":["Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/xception/xception_weights_tf_dim_ordering_tf_kernels_notop.h5\n","83683744/83683744 [==============================] - 1s 0us/step\n","Model: \"model\"\n","_________________________________________________________________\n"," Layer (type)                Output Shape              Param #   \n","=================================================================\n"," input_2 (InputLayer)        [(None, 256, 256, 3)]     0         \n","                                                                 \n"," sequential (Sequential)     (None, 256, 256, 3)       0         \n","                                                                 \n"," rescaling (Rescaling)       (None, 256, 256, 3)       0         \n","                                                                 \n"," xception (Functional)       (None, 8, 8, 2048)        20861480  \n","                                                                 \n"," global_average_pooling2d (G  (None, 2048)             0         \n"," lobalAveragePooling2D)                                          \n","                                                                 \n"," dropout (Dropout)           (None, 2048)              0         \n","                                                                 \n"," dense (Dense)               (None, 1)                 2049      \n","                                                                 \n","=================================================================\n","Total params: 20,863,529\n","Trainable params: 2,049\n","Non-trainable params: 20,861,480\n","_________________________________________________________________\n"]}],"source":["base_model = keras.applications.Xception(\n","    weights=\"imagenet\",  # Load weights pre-trained on ImageNet.\n","    input_shape=(256, 256, 3),\n","    include_top=False,\n",")  # Do not include the ImageNet classifier at the top.\n","\n","# Freeze the base_model\n","base_model.trainable = False\n","\n","# Create new model on top\n","inputs = keras.Input(shape=(256, 256, 3))\n","l = data_augmentation(inputs)  # Apply random data augmentation\n","\n","# Pre-trained Xception weights requires that input be scaled\n","# from (0, 255) to a range of (-1., +1.), the rescaling layer\n","# outputs: `(inputs * scale) + offset`\n","scale_layer = keras.layers.Rescaling(scale=1 / 127.5, offset=-1)\n","l = scale_layer(l)\n","\n","# The base model contains batchnorm layers. We want to keep them in inference mode\n","# when we unfreeze the base model for fine-tuning, so we make sure that the\n","# base_model is running in inference mode here.\n","l = base_model(l, training=False)\n","l = keras.layers.GlobalAveragePooling2D()(l)\n","l = keras.layers.Dropout(0.2)(l)  # Regularize with dropout\n","outputs = keras.layers.Dense(1)(l)\n","model = keras.Model(inputs, outputs)\n","\n","model.summary()"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"7YoDEGQn4RU-","executionInfo":{"status":"ok","timestamp":1665682820368,"user_tz":-330,"elapsed":77383,"user":{"displayName":"Akash Das","userId":"13248720121128963302"}},"outputId":"fb4ab86f-acac-452b-a951-f7662806384b"},"outputs":[{"output_type":"stream","name":"stdout","text":["Epoch 1/20\n","15/15 [==============================] - 16s 337ms/step - loss: 0.6959 - binary_accuracy: 0.4958 - val_loss: 0.6913 - val_binary_accuracy: 0.5167\n","Epoch 2/20\n","15/15 [==============================] - 3s 210ms/step - loss: 0.6926 - binary_accuracy: 0.4958 - val_loss: 0.6927 - val_binary_accuracy: 0.5167\n","Epoch 3/20\n","15/15 [==============================] - 3s 212ms/step - loss: 0.6871 - binary_accuracy: 0.4958 - val_loss: 0.6885 - val_binary_accuracy: 0.5167\n","Epoch 4/20\n","15/15 [==============================] - 3s 212ms/step - loss: 0.6903 - binary_accuracy: 0.4958 - val_loss: 0.6888 - val_binary_accuracy: 0.5167\n","Epoch 5/20\n","15/15 [==============================] - 3s 212ms/step - loss: 0.6844 - binary_accuracy: 0.4958 - val_loss: 0.6862 - val_binary_accuracy: 0.5167\n","Epoch 6/20\n","15/15 [==============================] - 3s 212ms/step - loss: 0.6886 - binary_accuracy: 0.4958 - val_loss: 0.6844 - val_binary_accuracy: 0.5167\n","Epoch 7/20\n","15/15 [==============================] - 3s 213ms/step - loss: 0.6933 - binary_accuracy: 0.4958 - val_loss: 0.6858 - val_binary_accuracy: 0.5167\n","Epoch 8/20\n","15/15 [==============================] - 3s 213ms/step - loss: 0.6888 - binary_accuracy: 0.4958 - val_loss: 0.6807 - val_binary_accuracy: 0.5167\n","Epoch 9/20\n","15/15 [==============================] - 3s 215ms/step - loss: 0.6865 - binary_accuracy: 0.4958 - val_loss: 0.6812 - val_binary_accuracy: 0.5167\n","Epoch 10/20\n","15/15 [==============================] - 3s 216ms/step - loss: 0.6847 - binary_accuracy: 0.4958 - val_loss: 0.6786 - val_binary_accuracy: 0.5167\n","Epoch 11/20\n","15/15 [==============================] - 3s 218ms/step - loss: 0.6833 - binary_accuracy: 0.4958 - val_loss: 0.6764 - val_binary_accuracy: 0.5167\n","Epoch 12/20\n","15/15 [==============================] - 3s 216ms/step - loss: 0.6800 - binary_accuracy: 0.4958 - val_loss: 0.6752 - val_binary_accuracy: 0.5167\n","Epoch 13/20\n","15/15 [==============================] - 3s 217ms/step - loss: 0.6776 - binary_accuracy: 0.4958 - val_loss: 0.6747 - val_binary_accuracy: 0.5167\n","Epoch 14/20\n","15/15 [==============================] - 3s 218ms/step - loss: 0.6739 - binary_accuracy: 0.4958 - val_loss: 0.6728 - val_binary_accuracy: 0.5167\n","Epoch 15/20\n","15/15 [==============================] - 3s 220ms/step - loss: 0.6764 - binary_accuracy: 0.4958 - val_loss: 0.6716 - val_binary_accuracy: 0.5167\n","Epoch 16/20\n","15/15 [==============================] - 3s 220ms/step - loss: 0.6729 - binary_accuracy: 0.4958 - val_loss: 0.6706 - val_binary_accuracy: 0.5167\n","Epoch 17/20\n","15/15 [==============================] - 3s 221ms/step - loss: 0.6705 - binary_accuracy: 0.4958 - val_loss: 0.6690 - val_binary_accuracy: 0.5167\n","Epoch 18/20\n","15/15 [==============================] - 3s 223ms/step - loss: 0.6729 - binary_accuracy: 0.4958 - val_loss: 0.6684 - val_binary_accuracy: 0.5167\n","Epoch 19/20\n","15/15 [==============================] - 3s 224ms/step - loss: 0.6717 - binary_accuracy: 0.4958 - val_loss: 0.6654 - val_binary_accuracy: 0.5167\n","Epoch 20/20\n","15/15 [==============================] - 3s 226ms/step - loss: 0.6667 - binary_accuracy: 0.4958 - val_loss: 0.6642 - val_binary_accuracy: 0.5167\n"]},{"output_type":"execute_result","data":{"text/plain":["<keras.callbacks.History at 0x7f098b1893d0>"]},"metadata":{},"execution_count":9}],"source":["model.compile(\n","    optimizer=keras.optimizers.Adam(),\n","    loss=keras.losses.BinaryCrossentropy(from_logits=True),\n","    metrics=[keras.metrics.BinaryAccuracy()],\n",")\n","\n","epochs = 20\n","model.fit(X_train, y_train, epochs=epochs, validation_data=(X_test, y_test))"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"h13jMHpI6b5X"},"outputs":[],"source":["def fine_tune(model, X_train, y_train, X_test, y_test, epochs = 10,learning_rate = 1e-5):\n","  base_model.trainable = True\n","  model.summary()\n","\n","  model.compile(\n","    optimizer=keras.optimizers.Adam(1e-5),  # Low learning rate\n","    loss=keras.losses.BinaryCrossentropy(from_logits=True),\n","    metrics=[keras.metrics.BinaryAccuracy()],\n","  )\n","\n","  epochs = 10\n","  model.fit(X_train, y_train, epochs=epochs, validation_data=(X_test, y_test))"]},{"cell_type":"code","source":["fine_tune(model,X_train, y_train, X_test, y_test)"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"Y2OqlrA6FFfb","executionInfo":{"status":"ok","timestamp":1665682958726,"user_tz":-330,"elapsed":138409,"user":{"displayName":"Akash Das","userId":"13248720121128963302"}},"outputId":"c72f29b7-a527-480c-f574-46a353f52a19"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["Model: \"model\"\n","_________________________________________________________________\n"," Layer (type)                Output Shape              Param #   \n","=================================================================\n"," input_2 (InputLayer)        [(None, 256, 256, 3)]     0         \n","                                                                 \n"," sequential (Sequential)     (None, 256, 256, 3)       0         \n","                                                                 \n"," rescaling (Rescaling)       (None, 256, 256, 3)       0         \n","                                                                 \n"," xception (Functional)       (None, 8, 8, 2048)        20861480  \n","                                                                 \n"," global_average_pooling2d (G  (None, 2048)             0         \n"," lobalAveragePooling2D)                                          \n","                                                                 \n"," dropout (Dropout)           (None, 2048)              0         \n","                                                                 \n"," dense (Dense)               (None, 1)                 2049      \n","                                                                 \n","=================================================================\n","Total params: 20,863,529\n","Trainable params: 20,809,001\n","Non-trainable params: 54,528\n","_________________________________________________________________\n","Epoch 1/10\n","15/15 [==============================] - 21s 906ms/step - loss: 0.6818 - binary_accuracy: 0.5500 - val_loss: 0.6491 - val_binary_accuracy: 0.5167\n","Epoch 2/10\n","15/15 [==============================] - 13s 866ms/step - loss: 0.6241 - binary_accuracy: 0.5938 - val_loss: 0.4958 - val_binary_accuracy: 0.5167\n","Epoch 3/10\n","15/15 [==============================] - 13s 877ms/step - loss: 0.5466 - binary_accuracy: 0.7167 - val_loss: 0.4151 - val_binary_accuracy: 0.8333\n","Epoch 4/10\n","15/15 [==============================] - 13s 877ms/step - loss: 0.5262 - binary_accuracy: 0.7708 - val_loss: 0.3865 - val_binary_accuracy: 0.8583\n","Epoch 5/10\n","15/15 [==============================] - 13s 858ms/step - loss: 0.4271 - binary_accuracy: 0.8167 - val_loss: 0.3197 - val_binary_accuracy: 0.8583\n","Epoch 6/10\n","15/15 [==============================] - 13s 858ms/step - loss: 0.4228 - binary_accuracy: 0.8167 - val_loss: 0.3365 - val_binary_accuracy: 0.8583\n","Epoch 7/10\n","15/15 [==============================] - 13s 856ms/step - loss: 0.4052 - binary_accuracy: 0.8167 - val_loss: 0.3062 - val_binary_accuracy: 0.8583\n","Epoch 8/10\n","15/15 [==============================] - 13s 862ms/step - loss: 0.4353 - binary_accuracy: 0.7979 - val_loss: 0.4394 - val_binary_accuracy: 0.8583\n","Epoch 9/10\n","15/15 [==============================] - 13s 864ms/step - loss: 0.4371 - binary_accuracy: 0.8083 - val_loss: 0.3046 - val_binary_accuracy: 0.8583\n","Epoch 10/10\n","15/15 [==============================] - 13s 866ms/step - loss: 0.4284 - binary_accuracy: 0.8104 - val_loss: 0.4588 - val_binary_accuracy: 0.8583\n"]}]},{"cell_type":"code","source":["# Scoring\n","predictions = (model.predict(X_test) > 0.5).astype(\"int32\")\n","predictions = predictions.reshape(1,-1)[0]\n","print(classification_report(y_test, predictions))\n","\n","# calculate roc curve\n","fpr, tpr, thresholds = roc_curve(y_test, predictions)\n","auc = roc_auc_score(y_test, predictions)\n","print('AUC: %.3f' % auc)"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"idisje0SbYD1","executionInfo":{"status":"ok","timestamp":1665684462839,"user_tz":-330,"elapsed":882,"user":{"displayName":"Akash Das","userId":"13248720121128963302"}},"outputId":"a95c5931-3217-48b7-aceb-dbedffdef3b6"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["4/4 [==============================] - 1s 169ms/step\n","              precision    recall  f1-score   support\n","\n","         0.0       0.78      1.00      0.88        62\n","         1.0       1.00      0.71      0.83        58\n","\n","    accuracy                           0.86       120\n","   macro avg       0.89      0.85      0.85       120\n","weighted avg       0.89      0.86      0.85       120\n","\n","AUC: 0.853\n"]}]}],"metadata":{"colab":{"provenance":[{"file_id":"19LDYToVnMi3jUgG0uxFwhYaHr3N_VcKa","timestamp":1665681687442}],"authorship_tag":"ABX9TyNBduUZ8ZjhaHp0oMuP+RH8"},"kernelspec":{"display_name":"Python 3","name":"python3"},"language_info":{"name":"python"},"accelerator":"GPU"},"nbformat":4,"nbformat_minor":0}