This is a sample implementation of a Deep Residual Network (ResNet) as introduced in Deep Residual Learning for Image Recognition. The original implementation using caffe can be found on github.
This uses the python version of the CIFAR-10 dataset. Some of the implementation details have been inspired by the DenseNet implementation and this blog post.
- Implement the ResNet-110 variant.
- There is no ReLU at the end of residual blocks.
- The SGD optimizer is used with a momentum of 0.9. The initial learning rate is 0.1, it gets divided by 10 at epochs 30.
- Batch size is 128.
- A dropout with a keep probability of 0.8 is used after each convolution. Convolutions don't use any bias.
The resulting accuracy is 88.2% on CIFAR-10 without data augmentation.
