Zhizheng Xiang (向 志政)

Actually, the output of NeRF `raw` is applied ReLU in this line: https://github.com/yenchenlin/nerf-pytorch/blob/223fe62d87d641e2bb0fb5bdbbcb6dad5efb2af3/run_nerf.py#L275

Try to add `downsample_train = 8` in `./configs/truck.txt` file. This helps you train the radiance field with 1/8-size images thus saving memory.

Adding the `--no_flip` option to the command line in my case results in 5s improvement in the inference phase. Although there is a minor reduction in image quality, the time...

I believe the `radii` parameter represents the radiance of a specific Gaussian distribution. If `radii > 0`, it indicates that this Gaussian distribution is "visible" from a particular viewpoint. Otherwise,...