Furthest distance for recognizing faces


I am trying to have my OpenMV cam mounted to a wall and want to determine information about people walking by and looking at it. For this reason, I am wondering:

1.) what is the maximum distance faces can be from the camera for face detection? Put another way, what is the minimum recognizable pixel size of faces?
2.) should face detection not work for this task, are there any recommendations about a different algorithm for OpenMV that could be used for this? (I could imagine something like posenet in Tensorflow would be suited to the task, but as this is my first time using OpenMV I’m not sure if there’s anything else well suited to recognizing people–rather than just faces, since they might be too small-- in a frame).

thank you in advance!

The faces must be close to the camera just a few feet away, or you could use one of the zoom lenses. Note this detector is trained on frontal faces, it won’t detect profile faces. Currently we don’t have any built-in function for people detection, however training a NN should work.

1 Like