Normalization of cam_coords in inverse_warp

[plaut]

Shouldn't cam_coords be normalized by its L2 norm before multiplying by the depth?
i.e., add:
cam_coords /= torch.norm(cam_coords, p=2, dim=1, keepdim=True)
before this line:

SfmLearner-Pytorch/inverse_warp.py

Line 40 in a866333

return cam_coords * depth.unsqueeze(1)

[ClementPinard]

depth is not distance to the camera ! It's actually the z coordinate, hence the third point coordinate set to 1.