about main table 1

[emjay73]

Thank you for sharing your code. I'm trying to reproduce the results in the main 1 table. Now I fully trained NeRF results (not 'ours' results) and all of the values are showing slightly worse than the values in the table. Following is the Test Set Results / Train Set Result / Result in the paper.

test		psnr	ssim	lpips	prd
nerf	flower	13.628	0.2909	0.7835	nan
nerf	fortress	15.618	0.4311	0.6794	nan
nerf	leaves	12.734	0.1451	0.7938	nan
nerf	trex	12.419	0.3743	0.6729	nan

train		psnr	ssim	lpips	prd
nerf	flower	13.062	0.2887	0.8028	nan
nerf	fortress	13.539	0.3868	0.7249	nan
nerf	leaves	12.38599	0.143	0.819662	nan
nerf	trex	12.58406	0.425573	0.692024	nan

paper		psnr	ssim	lpips	prd
nerf	flower	13.8	0.302	0.716	nan
nerf	fortress	16.3	0.524	0.445	nan
nerf	leaves	13.01	0.18	0.687	nan
nerf	trex	15.7	0.409	0.575	nan

Can I get a clue? Also, I wonder which dataset is used for the table among train/val/test set

[jeongyw12382]

Hmm... could you share the scripts that you've run to reproduce the results?

[emjay73]

I have not changed the script from the original one. Actually, the branch that I used is the default 'custom' branch. Could it be the reason? Or any version subtleties?

[jeongyw12382]

Hmm... I reported the result using codes in the main branch. Based on my memory, there is no difference between the custom and main branch while running the base code. I'll check the script and code tommorow and soon notice you the reason for this issue.

[jeongyw12382]

Sorry for being late. I'm currently preparing for another conference and the dew is coming. I will recheck this issue after my submission is completed.

[alex04072000]

Hi, I encountered the same issue of NaN when training with custom branch without COLMAP poses. Did you solve the problem?

[alien19]

@alex04072000 Have you solved it?