Need help with ResNet-on-Cifar10?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

126 Stars 83 Forks 42 Commits 2 Opened issues


Reimplementation ResNet on cifar10 with caffe

Services available


Need anything else?

Contributors list

No Data

ResNets experiments on cifar10 with caffe


    author = {Kaiming He and Xiangyu Zhang and Shaoqing Ren and Jian Sun},
    title = {Deep Residual Learning for Image Recognition},
    journal = {arXiv preprint arXiv:1512.03385},
    year = {2015}


This repository reimplements resnet experiments on cifar10 with caffe according to the paper "Deep Residual Learning for Image Recognition" ( The data augmentation means 4 pixels are padded on each side for every images during training. You can make datasets prepared by using the scripts.


The network structure is here(we only list the network of 20 depth):


First, you should make sure that your caffe is correctly installed. You can follow this blog's instructions if you use windows.(

for training

  caffe train -solver=solver.prototxt -gpu 0

for testing

  caffe test -model=res20_cifar_train_test.prototxt -weights=ResNet_20.caffemodel -iterations=100 -gpu 0


Result with data augmentation:


Repeated Reference
20 lyaers 91.94% 91.25%
32 layers 92.70% 92.49%
44 layers 93.01% 92.83%
56 layers 93.19% 93.03%
110 layers 93.56% 93.39%

notice:'Repeated' means reimplementation results and 'Reference' means result in the paper.We got even better results than the original paper

Compare result(without data augmentation):


PlainNet ResNet
20 lyaers 90.10% 91.74%
32 layers 86.96% 92.23%
44 layers 84.45% 92.67%
56 layers 85.26% 93.09%
110 layers X 93.27%

Blog address


We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.