deepcam-cn / yolov5-face

YOLO5Face: Why Reinventing a Face Detector (https://arxiv.org/abs/2105.12931) ECCV Workshops 2022)
GNU General Public License v3.0
2.01k stars 491 forks source link

训练自己数据集关键点时,因为遮挡未标造成的问题 #206

Open xs818818 opened 1 year ago

xs818818 commented 1 year ago

我在训练数据时,对于图片中未出现的关键点选择不标注,标签中设置为-1,但是训练后识别时,未标注的点会有相应的识别结果,请问在哪能看到对于关键点的识别置信度,我想对于低的不标注

derronqi commented 1 year ago

没有置信度,你可以看看yolov7-pose