There are some changes in both versions due to shift from windows to linux. The linux version uses multiprocessing for creating independent processes one each for each camera input, and for each camera source, it agin splits up the processes for different behaviours. However, because in windows "forking" of processes is not available, so the windows version doesn't use multiprocessing and rather uses multithreading.
Both versions, can be used for any number of behaviours and any number of input sources. (However I had tested for only two videos simultaneously).
The scripts/all_behaviours.py takes care of all behaviour functions we want to include. Currently, it only has one behaviour method faceMaskDetector which is used in the integrating_behaviours.py script. Other behaviour methods can also be included in the AllBehaviours class defined in all_behaviours.py script.
First, get the checkpoint file from here and place it in face-mask_model_info folder. Then place the two input videos in the videos/input folder.
The main script is integrating_behaviours.py. It takes two command line args, one each for two input videos.
Execute the below command in command prompt:
python integrating_behaviours.py --input1 path/to/video1 --input2 path/to/video2