densenet

This project includes:

• densenet.py, a keras implementation of DenseNet.

• cifar10-test.py, code to train the model on Cifar 10, also to test the trained model.

• vis_densenet.py, a demo to visualize the layers of the model, using keras-vis.

• testnet.py, a vgg-like model, a convnet model with bottle-neck layers and densenet-style skip connections.

References

1. Gao Huang, Zhuang Liu and K. Weinberger. Densely Connected Convolutional Networks. Arxiv. Github.

2. DenseNet Other implementations, especially, copied Christopher Masch’s code and fixed some issues for aligning to the original DenseNet paper.