I have performed face detection with both CNN and cascades but I want to go further into Facial Recognition system with complete onboard computation. I have worked on similar systems with OpenCV but I doubt the same can be implemented on my board. One possible approach I could take is training an object detection model for the same, but that is something I want to avoid. Are there any other approaches to get some robust system for this?
Hi, there’s a 10MB model that will run at about 7FPS on the upcoming OpenMV AE3 that just outputs a face hash given any face. When the AE3 comes out you’ll just be able to use this trivially.
For the RT1062, said model would run at under 0.1 FPS.