Current poolingLayer class does not support customizing padding. All computations are carried out under an assumption of zero padding. Some existing deep learning frameworks support padding for pooling layer. Therefore, we should consider getting padding as an input for pooling layer and consider it during computation.
beomyeol
commented
8 years ago
Current
poolingLayerclass does not support customizing padding. All computations are carried out under an assumption of zero padding. Some existing deep learning frameworks support padding for pooling layer. Therefore, we should consider getting padding as an input for pooling layer and consider it during computation.