Fashion MNIST classifier using Chainer

Requirements

$ python src/train.py -g 0

Options:

-g (--gpu) <int>: Optional
GPU device ID. Negative value indicates CPU (default: -1)
-m (--model) <model file path>: Optional
Model file path to load (default: None)
-b (--batch_size) <int>: Optional
Mini batch size (default: 128)
-p (--prefix) <str>: Optional
Prefix of saved model file. (default: fashion_<base size>)
--epoch <int>: Optional
Training epochs (default: 300)
--save-epoch <int>: Optional
Epoch interval to save model parameter file. 0 indicates model parameter is not saved at fixed intervals. Note that the best accuracy model is always saved even if this parameter is 0. (default: 0)
--optimizer <str>: Optional
Optimizer name (sgd or adam, default: sgd)
--lr <float>: Optional
Initial learning rate for SGD (default: 0.1)
--alpha <float>: Optional
Initial alpha for Adam (default: 0.001)
--lr-decay-iter <int>: Optional
Iteration interval to decay learning rate. Learning rate is decay to 1/10 at this intervals. (default: 100)
--weight-decay <float>: Optional
Weight decay (default: 0.0001)
--seed <int>: Optional
Random seed (default: 1)

VGG like neural network model
Data augmentation
- random crops
- random horizontal flips
- random erasing (see https://arxiv.org/abs/1708.04896)

MIT license