Question regarding efficiency

[Divadi]

Hi, thank you for the interesting work! I was wondering - what is the main source of the FPS difference between IA-SSD and 3DSSD? Looking at the general model flow of both methods, they seemed to have similar backbones with similar # of points. However, the FPS difference is 83 vs 28. I was wondering if you had any insight as to why.

[yifanzhang713]

Sorry for late reply @Divadi , thanks for your attention and it is a very insightful question. The reasons are twofold. The first is that the learning-based sampling strategy is much more effective and efficient than the Feature-FPS and FPS in 3DSSD. In addition, the super lightweight backbone design, which also benefits from sampling, can make more point clouds parallel, thus leading to much quicker inference.

Hope this can help you.