This is a cross platform python implementation of screen capture and object recognition using:
- mss library for screen capture. (this adds a lot of flexibility - you can use youtube, webcam or hand drawings for video source)
- opencv for neural net inference (in the example SSD mobile net v3) - for fast low latency inference
you can install mss for new verions of python using
python -m pip install -U --user mss
There is a version of cafe model detection in the cafe folder