maincold2 / Compact-3DGS

The official repository of Compact 3D Gaussian Representation for Radiance Field
Other
399 stars 21 forks source link

Viewing trained results #14

Closed alexhjkim closed 6 months ago

alexhjkim commented 6 months ago

Hello, thank you for your great work!

I read the README about the original SIBR interactive viewer of 3DGS not being able to support neural fields for view-dependent color. May I ask how you viewed the results of Compact-3DGS without using the viewer, and how you generated the videos of the results, as can be seen on the project page?

maincold2 commented 6 months ago

Now we have updated the code for rendering videos, which is originally from LightGaussians.

python render_video.py --source_path PATH/TO/DATASET --model_path PATH/TO/MODEL --skip_train --skip_test --video

We used trained models without neural fields for the visualization of ellipsoids in the paper (to visualize the effect of masking or R-VQ), and the final rendered results were with neural fields by using the python scripts.

Thanks!

alexhjkim commented 6 months ago

Thanks for the update!