sfu-gruvi-3dv / self-supervised-human-depth Goto Github PK

HI!
I wanted to generate a depth map aligned with original image using the output depth in npy. But it's not a depth map, more an approximated depth from a close 3D model (SIMPL)
Is it a limitation of the algorithm or should I access the depth before the last step?
Here is the superposition of the depth image (only 1 z) and original image we can see the problem.

Thank you!

Hello, can you please post a link to your paper?

The script creates npy files for depth values, but I suppos it's only usable with python, how can I convert these files to png ?

大哥！！请问论文里的定性实验展示的是mesh吗？如果是的话如何从深度图转换到mesh呢？
还有Tang等人的实验，他的github项目中提供的预训练模型下载链接已经失效了，我可以从你这里得到他们？