Dataset is split into 229,981 images for training, 12,691 images for validation and 12,818 images for testing. An ImageNet pretrained VGG-16 is used as the base network. Should be easy to plug in the ...