←back to thread

Show HN: Real-Time Gaussian Splatting

(github.com)
142 points markisus | 3 comments | 15 May 25 13:26 UTC | HN request time: 0.622s | source

LiveSplat is a system for turning RGBD camera streams into Gaussian splat scenes in real-time. The system works by passing all the RGBD frames into a feed forward neural net that outputs the current scene as Gaussian splats. These splats are then rendered in real-time. I've put together a demo video at the link above.
1. mandeepj ◴[15 May 25 14:14 UTC] No.43995314[source]▶
Another implementation of splat https://github.com/NVlabs/InstantSplat
replies(2): >>43995452 #>>43995479 #
2. jasonjmcghee ◴[15 May 25 14:25 UTC] No.43995452[source]▶
The quality is better, no doubt, but this method (from the paper) takes on the order of 10-45s depending on input from their table. Which is much better than 10 minutes etc.

That being said, afaict OP's method is 1000x faster, at 33ms.

3. markisus ◴[15 May 25 14:28 UTC] No.43995479[source]▶
Note that the method you linked is "Splatting in Seconds" where as real-time requires splatting in tens of milliseconds.

I'm also following this work https://guanjunwu.github.io/4dgs/ which produces temporal Gaussian splats but takes at least half an hour to learn the scene.