Problems with the quality of the detector components (facenet, faciallandmarks, gazenet, emotionnet)

Accelerated Computing Intelligent Video Analytics TAO Toolkit
21

Hello!

Could you specify in more detail which parameters and in which files can be edited to change the behavior of the model, please?

I use an increased threshold value in the file “deepstream_tao_apps/configs/facial_tao
/config_infer_primary_facenet.txt”, which you once advised me:

[class-attrs-all]
#pre-cluster-threshold=0.2
pre-cluster-threshold=0.6