Face detection not working as expected

[razvanphp]

Hello,

Thank you for your work and sharing it, really appreciated.

I've tried it for accurate face detection, but the results are not as good as I would expect. Some parts are not included in the mask and limiting to "[semnatic] face only" also does not do a great job, bleeding a lot.

Any idea how to improve this? Fine-tuning on some face datasets would help?

Here are some samples:

[CoderZhangYx]

Hi, you may try to prompt the model with "[semantic] human face", which is our employed datasets' annotation.
Besides, if you want super accurate segmentation masks, our model may not serve well because we focus more on universal tasks. You may use part-segmentation datasets like humanparsing, PartImageNet and pascal-part to finetune the model, or simply apply other strong specialist models.

[CoderZhangYx]

Furthermore, if you want to include human hair, try to prompt with "[semantic] head"