VGG-19 variant is just slightly different from the VGG-16 architecture, with just addition
of 3 convolution layers, but number of parameters were same.
In VGG-16 architecture, the fully connected layers added on five convolution blocks
contributed to the majority of the millions of parameters.
0
VGG-16 model looks simple as layers are added sequentially with uniformity but it was
quite complex to train as the number of parameters were in millions.
All of the above.
Fig: 1