why hrnet use 1/4 downsample in the begin of the net?

HRNet / HRNet-Semantic-Segmentation

The OCR approach is rephrased as Segmentation Transformer: https://arxiv.org/abs/1909.11065. This is an official implementation of semantic segmentation for HRNet. https://arxiv.org/abs/1908.07919

Other

3.13k stars 686 forks source link

why hrnet use 1/4 downsample in the begin of the net? #188

Open JoyLuo opened 3 years ago

JoyLuo commented 3 years ago

the output size of hrnet is 1/4 of input, because the begin of the net use conv with 2 stride 2 times if not use downsample, the result may be more better

yunhuang1241 commented 3 years ago

In paper the author did not explain why, but i suppose: 1.GPU memory limit 2.Computation cost