What is the optimal frame resolution and size for detectface method?

[jboundaries]

Found lagging if frame size goes beyond 100KB. Face detection accuracy affects if resolution comes down. What could be the optimal frame resolution and size for the face detect method?

https://stackoverflow.com/questions/57812858/mtcnn-face-detection-alignment-lagging-in-ip-camera-convention-behind-opencv-cv

https://stackoverflow.com/questions/57852160/how-to-scale-down-physical-size-of-image-in-python-opencv?noredirect=1&lq=1

related questions.