Hello!
We are working on counting targets in a frame, captured by a camera.
We are more interested in accuracy rather than speed (1 FPS would even be enough); so, we started implementing with Pytorch a Faster-RCNN on a ResNet50 backend.
However, I see no use of such a model on a Raspberry Pi, is it because it is not sufficiently light?
Which one would you suggest to use, on condition that we don't lose too much accuracy.
PS Pytorch mentions Weights Quantization: is it mandatory to let a NN work on Raspberry, or is it just recommended?
Thanks!
Re: Best Neural Network for Object Detection on Raspberry Pi 4
This is something similar to what you want to do. This is a project on Image Processing. It estimates the size of a crowd.
https://www.theengineeringprojects.com/ ... -pi-4.html
https://www.theengineeringprojects.com/ ... -pi-4.html