ResNet-32 is a convolutional neural network backbone that is based off alternative ResNet networks such as ResNet-34, ResNet-50, and ResNet-101. As its name implies, ResNet-32 is has 32 layers. It addresses the problem of vanishing gradient with the identity shortcut connection that skips one or more layers. The ResNet backbone can be ported into many applications including image classification as it is used here. This implementation of ResNet-32 is created with fastai, a low code deep learning framework.
ResNet-32's Architecture is largely inspired by the architecture of ResNet-34. Below, on the right-hand side, is Resnet34's architecture where the 34 layers and the residuals from one layer to another are visualized.
ResNet-32 Original Repository: https://github.com/verrannt/Tutorials