ResNets provide high classification accuracy while being parameter-efficient. Their building blocks allow very deep architectures to be built. The ResNet authors tested architectures with more than 1000 layers deep.
You can find more about ResNet architecture at:
- The Deep Residual Learning for Image Recognition paper.
- CAI Identity Shortcut Connection Example.
- Understanding and visualizing ResNets.
This folder contains 2 source code examples:
- ResNet-20: closely implements the ResNet-20 architecture.
- CaiResNet-20: implements a modified ResNet-20 that takes adantage of CAI specific layers.
There is also a web server source code example ready to be used.