Thanks for the information regarding the ESP32-AI-Thinker board’s inability to run Face Recognition.
I’ll use it to do snapshots!
This is not so much a question as information regarding Face Detection/Recognition on an ESP32 board with the S3 chip.
I purchased a FREENOVE ESP32-S3-WROOM-1 board. I was able to run the CameraWebServer program from the Examples->ESP->Camera repository.
Here is what I observed as the program’s behavior.
When I select Face Detection and I’m not too far away from the camera AND I remain reasonably still, I see a YELLOW box around my face. I have no idea what is happening beyond seeing the yellow box!
When I select Face Recognition along with the distance and stillness requirements, I see a GREEN box around my face along with what appears to be a matrix specification such as ID[n] : followed by what looks like a time specification. It comes and goes. The ‘n’ changes in what appears to be a random fashion such as 1, 7, 4, etc.
I was even able to observe (in RED) a message of “Intruder Alert”.
However, all of this is completely chaotic — I have no way of deciding if anything actually works. Part of the problem is I cannot find a written explanation of how the program should behave as a result of me clicking on the buttons!!
Lacking a video showing the operation of the program or a written explanation of it’s behavior, I can only conclude it either doesn’t actually work as intended or I’m too far from my router and there are too many interruptions in the data flow.
I guess I’m going to have to use my laptop for the program and sit right beside the router.
One of the things I can’t figure out is how do I know when the Enroll Face phase is complete, and I can then do a Face Recognition operation?
Also, where does it save the Face it is supposed to later Recognize? Does it require the SD-CARD or is it in memory?
I’m using the built-in antenna, because I’m slightly terrified of trying to unsolder the shorting jumper and resoldering it to the other position so I can use my external antenna.
I began by saying this is not so much a question — it appears I was wrong!
That’s actually how the program works.
When you get the intruder alert it means it couldn’t recognize your face (compared with the one that was enrolled).
The program is a bit slow. But it can be improved if you are closer to your router and have a good wi-fi signal.
However, you need to ake into account that face recognition and detection is a lot of processing for the ESP32. So, it will always be a bit slow.