Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

NeRF 9.5h results

See original GitHub issue

Which implementation do you use to generate these results? They seem much worse than the results reported in the paper and the results I got with my implementation.

Screenshot from 2021-07-07 00-18-45

Here’s my result with horns, PSNR reaches already 31.6 after 9.5h (reaches 25.91 after 22mins btw). I believe there should be some simplification in the implementation that you adopt, because I implement almost the same way as the original paper does, and adopt the same hyperparameters.

Issue Analytics

State:
Created 2 years ago
Comments:11 (6 by maintainers)

Top GitHub Comments

1reaction

apchenstucommented, Jul 12, 2021

recently I found the NDC ray sampling influent the quality of LLLFF a lot, so I am re-evaluating them and also IBRNet with their official code. I will update the result in our revision.

0reactions

apchenstucommented, Jul 13, 2021

Oh, the statement refers to the fine-tuning stage and the fine-tuning pairing file is in the ‘configs/pairs.th’, thanks~

Top Results From Across the Web

MVSNeRF: Fast Generalizable Radiance Field ... - Anpei Chen

We present MVSNeRF, a novel neural rendering approach that can efficiently reconstruct neural radiance fields for view synthesis.

MVSNeRF: Fast Generalizable Radiance Field ... - arXiv Vanity

We show results of our fine-tuning (top) and optimizing a NeRF [29] (bottom) ... We show our 15-min fine-tuning results and NeRF's [29]...

Neural Rays for Occlusion-aware Image-based Rendering

Paper tables with annotated results for Neural Rays for Occlusion-aware Image-based Rendering. ... NeRF [mildenhall2020nerf], 200k, ∼9.5h, 30.27.

Neural Rays for Occlusion-aware Image-based ... - arXiv

views, which achieve photo-realistic rendering results but typically have a limited renderable ... ilar quality as NeRF [33] with long training time (9.5h),....

Neural Rays for Occlusion-Aware Image ... - CVF Open Access

The results show that our method is able to be finetuned limited time (32min) to achieve sim- ilar quality as NeRF [31] with...