Fetching the paper…

Finet: Using Fine-grained Batch Normalization to Train Light-weight Neural Networks · Around