deepinsight / deepinsight/insightface
How to fine-tune ArcFace on a new dataset?
Open
- Dominant language
- Python
- Stars
- 29.7k
- Forks
- 6.1k
- PR merge metrics
- No merged PRs in 30d
Description
Hi, Thanks for this awesome work. I'm trying to use ArcFace on FFHQ. However, it seems there's domain gap so the running mean and variance does not directly fit to FFHQ. Has anyone tried to fine-tune ArcFace on other datasets and how to handle the BatchNorm layers?
I also tried to freeze the BatchNorm layers, which unfortunately leads to NaN.
Contributor guide
No contributing guide indexed for this repository
Assessment
This issue has not been assessed yet.