aipixel / GPS-Gaussian

[CVPR 2024 Highlight] The official repo for “GPS-Gaussian: Generalizable Pixel-wise 3D Gaussian Splatting for Real-time Human Novel View Synthesis”

Home Page:https://shunyuanzheng.github.io/GPS-Gaussian

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

why the pipeline in this paper can only reconstruct human body.

liuyaoli12345 opened this issue · comments

I'm new to these area and I've just read the paper. I'm wondering why the pipeline in this paper can only reconstruct human body. Which component introduces this limitation to the pipeline?
I'm terribly sorry if this question is stupid. I'm a newbie.

Hi, the framework of GPS-Gaussian is not limited to Human NVS. However, the real-time efficiency is highlighted in our method. Validation on static objects cannot showcase the main superiority of our method while NVS on moving objects is not a general application. So our method can be widely used in Human-centered scenarios such as holographic communication systems. Another reason is that we need ground-truth depth for training, many widely used NVS datasets (e.g. LLFF, DTU, etc.) do not have GT depth.