Hi, Thank you for your amazing work. I once conducted some research on unsup3d, and found the generated normals always have backgrounds, since they are constructed from depth maps. You also pointed out this in Figure 10 in your appendix.
Those normal backgrounds cause many troubles in the visualization especially in non-face objects, causing the result normals more like an improved form of depth map, as shown in Figure 10 in your appendix.
However, in all other sample figures in your paper such normal backgrounds are all removed. Could you please tell me how do you realized this, and where can I find the related code.
@JiuTongBro We use an off-the-shelf scene parsing model PSPNet to parse the foreground and background. The code is at the parse_mask function in gan2shape/model.py (line 386).
Hi, Thank you for your amazing work. I once conducted some research on unsup3d, and found the generated normals always have backgrounds, since they are constructed from depth maps. You also pointed out this in Figure 10 in your appendix.
Those normal backgrounds cause many troubles in the visualization especially in non-face objects, causing the result normals more like an improved form of depth map, as shown in Figure 10 in your appendix.
However, in all other sample figures in your paper such normal backgrounds are all removed. Could you please tell me how do you realized this, and where can I find the related code.