why not multiply ray step size when volume rendering.

[zhuixunforever]

Hi, thanks for this great code. I have a puzzle, in your code, https://github.com/apchenstu/mvsnerf/blob/6e0111188ff92b13229482a95a2fb4eaddbc2971/renderer.py#L22

the 'sigma' not multiply step size 'dist', which is different in Ep(8) in your paper.

By the way ,the origin NeRF: https://github.com/yenchenlin/nerf-pytorch/blob/62da0b218b41573535a59ac1a38e9aeb840385a2/run_nerf.py#L275 raw2alpha = lambda raw, dists, act_fn=F.relu: 1.-torch.exp(-act_fn(raw)*dists)

https://github.com/yenchenlin/nerf-pytorch/blob/62da0b218b41573535a59ac1a38e9aeb840385a2/run_nerf.py#L293 alpha = raw2alpha(raw[...,3] + noise, dists) # [N_rays, N_samples]

Any help will be greatly appreciated!

[apchenstu]

Hi, I have checked that Ep(8) seems doesn't contain "dist" also. The trick of removing the marching size comes from the IBRNet for better generalizability, but it may not be correct for general volume rendering, hope this can help you.

[zhuixunforever]

Sorry for the late reply.

According to my understanding, delta in the second formula of Ep(8) is 'dist'. Am I getting this wrong?

[apchenstu]

Sorry for the late reply.

According to my understanding, delta in the second formula of Ep(8) is 'dist'. Am I getting this wrong?

maybe you are using our initial version, you can download the newest version from arxiv

[MELANCHOLY828]

Do you solve the problem?I have the same question

[MELANCHOLY828]

ok!I get!