Implementation of Pyramidal Residual Networks by chainer
git clone https://github.com/nutszebra/pyramidal_residual_networks.git
cd pyramidal_residual_networks
git submodule init
git submodule update
python main.py -g 0
All hyperparameters and network architecture are the same as in [1] except for data-augmentation.
- Data augmentation
Train: Pictures are randomly resized in the range of [32, 36], then 32x32 patches are extracted randomly and are normalized locally. Horizontal flipping is applied with 0.5 probability.
Test: Pictures are resized to 32x32, then they are normalized locally. Single image test is used to calculate total accuracy.
| network | alpha | depth | total accuracy (%) |
|---|---|---|---|
| Pyramidal Residual Networks [1] | 270 | 110 | 96.23 |
| my implementation | 270 | 110 | 95.9 |
Deep Pyramidal Residual Networks [1]