search

irfanICMLL / structure_knowledge_distillation

The official code for the paper 'Structured Knowledge Distillation for Semantic Segmentation'. (CVPR 2019 ORAL) and extension to other tasks.
BSD 2-Clause "Simplified" License
702 stars 104 forks source link

The paper use average pooling to aggregate β × C features in one node to be 1 × C, but the code shows max pooling. #46

Open aye0804 opened 4 years ago

aye0804 commented 4 years ago

The paper use average pooling to aggregate β × C features in one node to be 1 × C, but the code shows max pooling: maxpool = nn.MaxPool2d(kernel_size=(patch_w, patch_h), stride=(patch_w, patch_h), padding=0, ceil_mode=True)

irfanICMLL commented 4 years ago

After the paper, we found that max-pooling can achieve better performance. We will change the paper version to the Arxiv.