October 2017
Beginner to intermediate
270 pages
7h
English
The main characteristic of the VGG network architecture is that it reduced the size of the convolutional filters to a simple 3 x 3 matrix and combined them in sequences, which was different to previous contenders, which had large kernel dimensions (up to 11 x 11). Paradoxically, the series of small convolutional weights amounted to a really large number of parameters (in the order of many millions), and so it had to be limited by a number of measures.

Read now
Unlock full access