search

speedinghzl / CCNet

CCNet: Criss-Cross Attention for Semantic Segmentation (TPAMI 2020 & ICCV 2019).
MIT License
1.42k stars 277 forks source link

Why didn't add self.INF(m_batchsize, height, width) to energy_W #122

Open Theia137 opened 3 days ago

Theia137 commented 3 days ago

energy_H = (torch.bmm(proj_query_H, proj_key_H)+self.INF(m_batchsize, height, width)).view(m_batchsize,width,height,height).permute(0,2,1,3) energy_W = torch.bmm(proj_query_W, proj_key_W).view(m_batchsize,height,width,width)

What's difference between energy_H, and energy_W that coder didn't add self.INF to energy_W?