search

Tencent / FaceDetection-DSFD

腾讯优图高精度双分支人脸检测器
Other
2.91k stars 728 forks source link

I notice that the visual threshold used in your work is very low. Wouldn't it result in a lot of false positives? #10

Closed yxchng closed 5 years ago

yxchng commented 5 years ago

I notice that the visual threshold used in your work is very low (0.1 in demo and 0.01 in WIDERFACE). Wouldn't it result in a lot of false positives?

I tested your demo and it indeed gives many false positives. Is it normal?

swordlidev commented 5 years ago

Hi @yxchng Yes, You are free to change the visual threshold in demo.py for different application. you can try 0.8 in demo.py to decrease false positives. However, visual threshold must be set 0.01 in widerface_val.py. Because mAP in evaluation is determined by both precision and recall.