search

jkulhanek / wild-gaussians

WildGaussians: 3D Gaussian Splatting In the Wild
https://wild-gaussians.github.io
Other
255 stars 20 forks source link

How to specify image resolution size #8

Closed santolina closed 5 days ago

santolina commented 1 month ago

Thank you very much for your great work.

BTW, Is there any parameter to explicitly specify image resolution like "--resolution 1920" in vanilla GS ?

jkulhanek commented 1 month ago

Hi, sorry, but at the moment there is no argument to specify the image resolution. You can copy a dataset and add an nb-info.json file with a custom downscale factor. Would this resolve your issue?

santolina commented 2 weeks ago

Thanks a lot for your comment. Using nb-info.json is sufficient to benchmark public datasets.

yuzewang1998 commented 1 week ago

Hi, could you please give more information about how to add such nb-info.json and use it? I can't find such config file in the code of Wild-Gaussians or NerfBaselines.

jkulhanek commented 1 week ago

Can you please take a look at the released NeRF on-the-go dataset? Each scene should contain nb-info.json so you can take a look at how it is defined. It basically defines the dataset's metadata. By setting downscale_loaded_factor=2 the images will be automatically downscaled by a factor of 2 on loading. Take a look at the relevant code: https://github.com/jkulhanek/nerfbaselines/blob/72a3aa8c55bca9fe190d76645fd7db710250661e/nerfbaselines/datasets/_common.py#L247C39-L247C62

yuzewang1998 commented 1 week ago

Thank you very much for your patience in explaining. I started to download only the PT dataset, so I didn't see the nb-info.json, I'm very sorry.

jkulhanek commented 5 days ago

Please take a look at the documentation https://jkulhanek.com/nerfbaselines/docs/using-custom-data.html