Inference time
hungsing92 opened this issue · 4 comments
hungsing92 commented
Hi,
Many thanks for your work.!
I use your lib to render a image, but it seems to be slow. I find many loops in your code. Can it speed up ? Or any suggestion?
Best,
SteveJunGao commented
We have an optimized version, please check it here: https://github.com/NVIDIAGameWorks/kaolin/tree/master/examples/renderers/DIB-R/graphics
hungsing92 commented
Thanks for your quick reply!
I have tried this version. It costs 200ms to render one picture, but tf_mesh_renderer just costs 20ms. It seems that there is some difference between your pipeline.
Best,
wenzhengchen commented
We re-write OpenGL by cuda in PyTorch, and that's why it is slower than directly calling OpenGL. We are preparing a faster version.
jli113 commented