This is an experimental implementation of Densely Connected Convolutional Networks using Chainer framework.

• Original paper: https://arxiv.org/abs/1608.06993
• Official implementation: https://github.com/liuzhuang13/DenseNet

• Chainer 2.0+
• CuPy 1.0+

Train a DenseNet on CIFAR-10 dataset using:

python code/train.py --gpu 0 -o result

Visualize training result using:

python code/visualize.py result/log -o result

• Model parameters: L = 40, k = 12
• Batch size: 64
• Dataset: C10 (CIFAR-10 dataset without data augmentation)