deepinsight / deepinsight/insightface
[SCRFD] inference time
Open
- Dominant language
- Python
- Stars
- 29.7k
- Forks
- 6.1k
- PR merge metrics
- No merged PRs in 30d
Description
I did some modifications on the SCRFD model and its performance failed my expectation.
I expanded 5 keypoints output to 68 and added a separate conv-bn module just like for bbox and cls_score.
Also I added a mask module to output the mask probability of each face.
The complexity is shown blow:
`ON VGA:
4.1717028 G
0.931747 M`
The benchmark inference time is approximately 13ms (2080ti).
Also I have doubled the FLOPs comparing to original SCRFD_2.5G, but is this inference time reasonable?
Contributor guide
No contributing guide indexed for this repository
Assessment
This issue has not been assessed yet.