Open ptc-lexvandersluijs opened 3 months ago
We begin by optimizing a neus as the first step to intilize the shape. Then, we fix the geometry and primarily focus on optimizing the appearance.
To start, run python launch.py --config configs/neus-colmap.yaml --gpu 0 --train dataset.root_dir=$1
first, followed by python launch.py --config configs/bakedsdf-colmap.yaml --gpu 0 --train dataset.root_dir=$1 \ --resume_weights_only --resume latest
Oh wow, now I get it! I thought those two commands were simply variants "if you want to train with NeuS you use this, and if you want to train with BakedSDF you use the other", but now I understand that one needs to be run after the other! Oh man. Thank you very much, I will try again, armed with this new knowledge! :-)
Hi,
This works looks extremely interesting for a small hobby 3D scanner project I'm working on, but I can't get it to work. From the paper and the project site, the results look more promising than any other AI multi-view reconstruction research I've seen, and I've tried almost all open source variants by now I believe. So I'm hoping that someone is able to give some hints to get it to work.
When I try to run with the bakedsdf-colmap configuration and the garden dataset, using CUDA 12.1 and PyTorch 2.3.0 OR CUDA 11.8 and PyTorch 2.1.0, the system loads and runs, and then doesn't work at all: the images in the 'save' folder look the same at iteration 5000 and iteration 170000 (see attachments below).
This is on Ubuntu 20.04 (on dual-boot laptop, so not via WSL) with an RTX 3080 16 GB.
I do see these warnings after starting:
During training, the loss stays at approx. 0.6 and the PSNR at 11.80 Could the non-convergence be caused by some incompatibility in the Torch version? Or am I using an incorrect configuration file? Which version of PyTorch was used by the authors?
Any help would be greatly appreciated.
Below are the images from the 'save' directory, the first is saved at 10,000 iterations, the second at 170,000.