Questions about face recognition

I’m using seeed’s wio terminal and grove AI vision camera module. Currently working with circuitpython on the wio.

I’ve followed face detection tutorials but is there a way to train the model for facial recognition? Ideally I’d like to run the code on in the wio terminal, whiich triggers the grove vision for to scan my face and either match or label as unknown. Can I train a model to achieve this on Edge Impulse?

Hi @sd2324,

You can train an object detection model (e.g. MobileNetv2-SSD) to identify individual faces. However, that model is too large to run in a Wio Terminal. You will need something like a Raspberry Pi for that. You can use our FOMO algorithm to identify that a face (i.e. any face) exists in a general area of the image, but it likely will not be able to uniquely identify different faces.

As far as I know, CircuitPython does not have support for TensorFlow Lite Micro, so you would need to write your code in C++ (or Arduino) if you want to run a model trained in Edge Impulse on your Wio Terminal.