1adrianb / 1adrianb/face-alignment

blazeface No face detected in FIRST source image

未关闭
#364 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看
主要语言
Python
星标
7.5k
派生
1.4k
PR 合并指标
30 天内没有已合并 PR

描述

Hi, I hope this is the right place to post this question.

I'm trying to use Blazeface in a Liveportrait workflow in the Liveportrait LoadFaceAlignment Cropper Node. I had it working a few weeks ago but had to uninstall and reinstall. Now Blazeface & Blazeface Back Camera both give the error message: blazeface No face detected in FIRST source image

I was able to get SFD to work, so my first question is if SFD is just as good. If it is, please let me know and ignore the rest.

I seem to remember that there was a setting for "confidence" that would fix this problem, but I looked through every node 3 times and can't find it.

I also read somewhere that blazeface uses 6 keypoints. 2 eyes, 2 ears, nose, & mouth. The subject has her ears covered by long hair. Could this cause the problem?

Those are the causes I could think of. If someone can give me some ideas I'd really appreciate it.

贡献指南

这个仓库没有索引到贡献指南

调研方向

The issue is about Blazeface failing to detect a face in the first source image within a Liveportrait workflow. Check the Blazeface model implementation and its confidence threshold settings. Look at the face-alignment library's source code for the Blazeface detector, possibly in face_alignment/detection/blazeface.py. Review how keypoints (eyes, ears, nose, mouth) are used and if hair covering ears affects detection. Compare with SFD detector to understand differences.

由索引模型根据 Issue 内容生成。

评估

技术栈
python, pytorch
领域
ai, computer-vision
Issue 类型
缺陷
难度
4/5
预计耗时
3-5 天
活跃度
停滞
描述清晰度
基本清楚
新手友好度
35/100

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。