search

cvlab-kaist / RAIN-GS

Code for "Relaxing Accurate Initialization Constraint for 3D Gaussian Splatting" by Jaewoo Jung, Jisang Han, Honggyu An, Jiwon Kang, Seonghoon Park, and Seungryong Kim
https://ku-cvlab.github.io/RAIN-GS
MIT License
290 stars 16 forks source link

Some questions about the Progressive Gaussian low-pass filter control and ABE-Split #15

Open JIANG-CX opened 3 months ago

JIANG-CX commented 3 months ago

Wonderful work!

I have some questions about the effectiveness of the Progressive Gaussian low-pass filter control and the ABE-Split strategy:

  1. Since you adopt SLV random initialization, each Gaussian already has a large covariance and covers a larger area. Why is it still necessary to further enlarge their coverage area using the low-pass filter?
  2. I'm a bit confused about the ABE-Split strategy. It seems that you clone a new Gaussian to a distant area, but how do you ensure that this cloned Gaussian is positioned appropriately?

Thanks.