awesome-semantic-segmentation-pytorch
awesome-semantic-segmentation-pytorch copied to clipboard
danet
attention_new = torch.max(attention, dim=-1, keepdim=True)[0].expand_as(attention) - attention I don't understand
same