This is a template repository for running experiments in computer vision classification problems using deep neural networks. Included are many of the standard architectures, including ResNet, ResNeXt, VGG and AllCNN. The main starting point is experiment_template/, which contains a barebones script for training a model on one of the common image classification datasets. You can copy this folder and modify it as necessary to suit your needs.

The folder utils/ has some useful tools for dissecting a trained model (including a plotting script), as well as a dispatcher tool for running many experiments sequentially.