search

ShichenLiu / SoftRas

Project page of paper "Soft Rasterizer: A Differentiable Renderer for Image-based 3D Reasoning"
MIT License
1.24k stars 156 forks source link

Question about the sigma value #52

Closed lijing1996 closed 3 years ago

lijing1996 commented 4 years ago

Thanks for sharing your nice work!

My question is why the sigma value in test.py is larger than train.py. In train.py:

https://github.com/ShichenLiu/SoftRas/blob/f6442793d0b3a4e02e76c2dd3a0209e1ac6c278b/examples/recon/train.py#L36

And it is decayed to be smaller during training. But in test.py:

https://github.com/ShichenLiu/SoftRas/blob/f6442793d0b3a4e02e76c2dd3a0209e1ac6c278b/examples/recon/test.py#L27

Maybe the sigma value in test.py should be smaller than in train.py. Thanks.

ShichenLiu commented 3 years ago

In fact, the sigma value is useless in the test.py. It is for providing gradients for the inverse rendering process.