Does anyone have a nice code for real-time detection on Raspberry Pi 4?

[HaroldSP]

Hello everyone!

Does anyone have a nice code for real-time detection on Raspberry Pi 4?

Right now I use this and it's working:

./darknet detector demo cfg/coco.data cfg/yolov3-tiny.cfg yolov3-tiny.weights

But very slow - it shows only 0.5 FPS.

Another thing is that I do not quite understand how to use it further - simply is that I can't find the executable file to edit it.

let's say I want to make my robot stop when it sees a cat from a rpi cam video stream.

I mean it is not so hard I guess to write it from scratch, but I am quite sure that it's already been done.

Thanks, everyone.

[ghost]

Hello, unfortunately there are still many unresolved issues with this repository. I couldn't set darknet-nnpack on Raspberry Pi 3. I suggest you to check an alternative solution.

1. You train the model and convert to tflite using tensorflow-yolov4-tflite
2. On Raspberry PI, you only setup the Tensorflow Lite interpreter
3. You write a script that loads data and processes network results in Python.

In this way I can get 1 FPS on Raspberry Pi 3b+ with yolo3 tiny.

I would be happy to get to know other ways.