wanmeihuali
wanmeihuali
See https://github.com/wanmeihuali/taichi_3d_gaussian_splatting/issues/151
…use the occlusion shall be more correct with depth in ray.
the article(https://tellusim.com/07_hello_splatting/) claims that removing shared memory and inference 8 pixels per thread have a good acceleration on inference. May be we can try that. Let's first try adjust the...