You can use different weights at the cost of decreased accuracy in detection

[naveenmula]

You can use different weights at the cost of decreased accuracy in detection

Originally posted by @hypadr1v3 in https://github.com/sthanhng/yoloface/issues/4#issuecomment-651102477

[naveenmula]

Hello, I am using the same weights for my master thesis, and it is showing good accuracy, a less false positive, less false-negative rate that I can notice with the subjective evaluation. Could you send me training and testing curves (Confusion Matrix)? Could you please give an accuracy of this model that I can mention in my thesis.

[hypadr1v3]

Hi, sorry for not replying earlier. This code does not have any training code and thus only @sthanhng can give you this although I would recommend to go through the YOLOv3 paper and use their mAP etc. Although if you really need such values, then I suggest reproducing what this code did by either implementing YOLOv3 yourself or using another code base that can also train and then use the Wider Faces dataset to train it and get results.