How to overcome the underfitting of 4DGS

[liyw420]

Hello! Thanks a lot for your great work!

Recently I tried your model on other Dynamic View Synthesis Dataset, like the Google Immersive Dataset. However, I found that your model sometimes cannot capture the detailed textures of the scene. Both rendered depths and images are lack textures with large and blurry artifacts. Here is an example (people below are blurry):

GT image: Rendered image: Rendered depth:

I was wondering if you ever meet some problems like this, and do you know any method to solve it? Thank you!

[liyw420]

The input point clouds are obtained from COLMAP.