search

modelscope / 3D-Speaker

A Repository for Single- and Multi-modal Speaker Verification, Speaker Recognition and Speaker Diarization
Apache License 2.0
1.02k stars 89 forks source link

关于人脸相关模型输入通道的问题。 #95

Closed hao-qiang closed 4 months ago

hao-qiang commented 4 months ago

https://github.com/alibaba-damo-academy/3D-Speaker/blob/4590d1cbb89fd9176a9d57573969b1b8790bf68d/egs/3dspeaker/speaker-diarization/local/vision_tools/face_quality_assessment.py#L18-L20

在人脸质量评估和人脸特征提取模型的输入数据处理的地方，通道被转换了两次，相当于没有变，请问这两个模型的输入通道是RGB还是BGR。

wanghuii1 commented 4 months ago

感谢指正。此处应该是一个错误，对比模型源码，网络输入应是RGB。不过经过对比，这个错误产生的误差非常小。目前代码已经更正，并对输入格式进行了注释。