October 2017
Beginner to intermediate
270 pages
7h
English
This new architecture appeared in December 2015 (more or less the same time as Inception V3), and it had a simple but novel idea—not only should the output of each constitutional layer be used but the architecture should also combine the output of the layer with the original input.
In the following diagram, we observe a simplified view of one of the ResNet modules. It clearly shows the sum operation at the end of the convolutional series, and a final ReLU operation:

The convolutional part of the module includes a feature reduction from 256 to 64 values, a 3 x 3 filter layer maintaining the feature numbers, and a ...
Read now
Unlock full access