shuokay / resnet

Deep Residual Learning for Image Recognition, http://arxiv.org/abs/1512.03385

85 stars 39 forks source link

cnn deep-learning deep-residual-learning

readme

Deep Residual Net

Example code for Deep Residual Learning for Image Recognition

Run this script by python resnet-small.py for 100 epochs get a train accuracy around 89.47% and validation accuracy around 85.95%
Then change the learning rate to 0.01, running this training from 100th epoch for 50 iterations, and get a train accuracy around 98.72% and test accuracy around 89.77%

Differences to the Paper

1*1 convolution operators are used for increasing dimensions.
This is a small residual net consists of 52 layers(can change to 20, 32, 44 layers by changing n in ResidualSymbol to 3, 5, 7)
Using mxnet default data augmentation options include center crop (instead of random crop) and random mirror, no paddings on raw image data and the input image size is 28*28(instead of 32*32).