Jul 25, 2017 · 1 min read
You can ignore the object_detection_multilayer.py file. It’s an earlier idea from me where I experimented with multiple layers on OpenCV so that one layer would be the original input (videostream) and the other layer would be the detected classes/boxes. The blend_non_transparent basically removes the “background” from the detected classes/boxes. The idea actually worked and performed quite well so that I left the implementation in my Git for future references. Hope this helps!
