[3D CV 연구] 3DGS The number of points is decreasing when training large-scale scenes

large-scale scenes에 대해서 gaussian의 수가 크게 줄어드는 경우, 어떻게 해결하나요?

Large gaussian들이 culling되는 것을 comment out 한 다음 결과를 봐보면 됩니다.

The algorithm has the potential for a draining effect due to the culling of large gaussians.

I will see if i can make a quickfix for it, but in the meantime, you could go and comment this line

prune_mask = torch.logical_or(torch.logical_or(prune_mask, big_points_vs), big_points_ws)

in gaussian_model.py and that should prevent it