I implemented the "face detection and recognition" features in the ESP32 camera some time ago (and they are still present in the documentation) but I removed them since they were just a demonstration of the capabilities of the camera and not a real-world use case. The ESP32-S3 should be better as it contains specific AI processor instructions that are supposed to improve performance on these particular applications. However, I do not think this will lead to significant improvements, particularly for your specific requirements. Probably Nvidia H/W cards are more appropriated.[Local Link Removed for Guests] wrote: [Local Link Removed for Guests]Mon Jan 06, 2025 1:16 pm
I also have S3 boards with camera, and it's not that I'm not interested in using Annex with them, it's just that I'd need functions that Annex will probably never support (like simultaneous detection of multiple objects).
I'm not even sure that what I have in mind will ever work with an ESP32.
You might want to look online to see if someone has developed a practical application using this camera and start from there.
You can find some inspiration here: https://github.com/espressif/esp-dl