Questions about FPS tested on A100?

MCG-NJU / SparseOcc

[ECCV 2024] Fully Sparse 3D Occupancy Prediction & RayIoU Evaluation Metric

https://arxiv.org/abs/2312.17118

Apache License 2.0

235 stars 16 forks source link

Questions about FPS tested on A100? #35

Closed harrylin-hyl closed 2 months ago

harrylin-hyl commented 3 months ago

Wonderful work! I also tested the speed of SparseOcc(8f) and FB-Occ on a A100 40G, but get some different results of the paper reported. FB-OCC: 10.0 FPS and SparseOcc(8f): 17.3 FPS in the paper, but we got FB-OCC: 7.0 FPS and SparseOcc(8f): 17.8 FPS. I want to know whether there are some test settings different, especially the test of FB-OCC (we also followed the open-source code)?

afterthat97 commented 3 months ago

Emmm I'm not sure :) Maybe I turned on fp16 when testing FBOcc... I have left OpenDriveLab and have no access to the previous data... Anyway, the speed of SparseOcc is OK, glad to see you reproduced it :)

harrylin-hyl commented 2 months ago

Thanks for your reply!