I read your comment about that face recognition in openMV not working in the real world.
but I want to program a face recognition system without using openMV examples;
I am using a tflite model for face detection which I downloaded from a reliable github repository but when I load it I get the following error:
My undergraduate project is about this and I don’t have a lot of time, do you think I can implement a facial recognition system with the board I bought or should I use something else like a raspberry pi?
Hi, face detection is easier and we can do that. Recognition is much harder. I suggest you use an online service with an API to dot this as it’s not something you do as a class project quickly.
Thanks.
So, I have to train a good performing tflite model for “object detection” and write a report about it. I tried to test several tflite models on board but my board failed to load them! Can you suggest a solution to build the model that run on openMV H7 board?
Use edge impulse to generate a firmware image and then load the firmware image. This allows you to run large models on the H7 R2. You cannot load the Tf lite file into RAM like on the H7 Plus because you don’t have as much in the R2.
I didn’t understand about ‘firmware image’. Actually, I’m going to improve the model currently used in face_detection.py example!
Considering that you said my board can’t support tflite model for facia detection, I think my job has become more difficult.
Can I generate a model with better performance using edge impulse?
Edge Impulse has an example that shows you how to build a firmware image for the OpenMV cam that has your model embedded in the firmware. This is required to get around the out of memory issue.
Please read their tutorial online about how to do this.
