search

dontLoveBugs / CSPN_monodepth

Unofficial Faster PyTorch implementation of Convolutional Spatial Propagation Network
67 stars 16 forks source link

Up projection layers #7

Open I3aer opened 3 years ago

I3aer commented 3 years ago

Hi,

Have you checked the implementations of Gudi_UpProj_Block and Gudi_UpProj_Block_Cat? The two blocks are not same as those depicted in the paper. In addition, there is an unused layer self.conv3 in the ResNet class.