Heavy machine learning models don't work everywhere, especially not on the edge where computing power is not as good. I have recently developed a simple and fast working procedure to achieve motion tracking in real time with raw Python and OpenCV.
The steps would involve grabbing frame by frame and calculating the absolute difference which is then blurred to reduce noise. The limitation is to set the contour area threshold to 9200 pixels which allows the software output spot the legitimate movement.
No API calls used. No excess work done. Only pure and fast execution.
Back to the code.
Top comments (0)