chulaihunde / chainer-cifar10

Various CNN models including Deep Residual Networks (ResNet) for CIFAR10 with Chainer (http://chainer.org)

chainer-cifar10

Requirement

• Chainer v1.5.0.2
• scikit-image 0.11.3
• scipy 0.16.0
• numpy 1.10. 1

Download & Construct Cifar10 Dataset

$ bash download.sh

Start Training

$ nohup python train.py &

Draw Loss Curve

$ python draw_loss.py --logfile log.txt --outfile log.jpg

Deep Residual Network (ResNet-110)

$ python dataset.py --whitening 0
$ python train.py --model models/ResNet.py --lr 0.1 --gpu 0

The test accuracy after 15 epochs is 0.9406 (error (%): 5.94). The test accuracy reported in the MSR paper (Deep Residual Learning for Image Recognition) is 0.9357 (error (%): 6.43) (see Table 6).

Resulting loss curve