This method can overcome the generalization problem between different cameras.
You don't need to repeat the training model anymore. Free your hands!!
You can look at our demo and try it on your own.
The demo's model only use 20 thousand to train and Face Detection using OpenCV Haar Cascades.So it has a great room for improvement!
TO DO
1.use more data to train the model
2.use FeatherNetB to train