Build a classifier to classify X-ray images diagnosis
https://nihcc.app.box.com/v/ChestXray-NIHCC/folder/37178474737 Only images_001.tar.gz, images_0012.tar.gz and images_003.tar.gz were used for this challenge. Those three compressed files contain 24999 images, which are in format of (1024, 1024, 3).
Download the data to directory 'images'
Resize each image into (224, 224, 3), and encode the labels, saved as "Resampled_data.h5".
Split the data into train, validation and test dataset, compute and save the bottlenexk features into "bottleneck_features_224_224_train.npy", "bottleneck_features_224_224_validation.npy" and "bottleneck_features_224_224_test.npy"
Build a fully convolutional layer to train the model
Show what the model learns by plotting class activation maps