Enhance VGG16 model

[bottydim]

The original paper states that "The training was regularised by weight decay (the L2 penalty multiplier set to 5 · 10−4) and dropout regularisation for the first two fully-connected layers (dropout ratio set to 0.5).

I think we have forgotten to add them:

    x = Dense(4096, activation='relu', name='fc1')(x)
    x = Dense(4096, activation='relu', name='fc2')(x)
    x = Dense(classes, activation='softmax', name='predictions')(x)