search

Anttwo / SuGaR

[CVPR 2024] Official PyTorch implementation of SuGaR: Surface-Aligned Gaussian Splatting for Efficient 3D Mesh Reconstruction and High-Quality Mesh Rendering
https://anttwo.github.io/sugar/
Other
2.21k stars 166 forks source link

Dataset error #220

Open zqdsdfg opened 1 week ago

zqdsdfg commented 1 week ago

Can I share training data sets? I train with the nerf_synthetic/lego dataset and keep reporting errors: "FileNotFoundError: [Errno 2] No such file or directory: '/home/zhang/3D/nerf_synthetic/lego/images/r_100.png' ", I checked the number of images in the data set and only r_0-r_99, I don't understand what's wrong?

Anttwo commented 1 week ago

Hello @zqdsdfg,

May I ask you when did you get the error during training? It would be very nice if you could post the logs/a screenshot of the console here.

The nerf_synthetic dataset has a different format than COLMAP datasets, that's probably why you get an error (as the code is focused on COLMAP-like real scenes). However, with a few adjustments to the loading scripts in the code, it's possible to run SuGaR on the nerf_synthetic scenes.

I don't have that much time right now, but I'll update the repo with some instructions to run the code on nerf_synthetic.

In the meantime, you can take a look at our more recent project Gaussian Frosting and check the ./Frosting/frosting_scene/cameras.py script, as it has the same architecture as SuGaR and is compatible with nerf_synthetic scenes: https://github.com/Anttwo/Frosting