It's a demonstration project,by detecting  image from web video camera . Before the image enters the neural network,
I do some image_processing with detecting multiple ROI windows' value to find out abnormal area. Due to background 
interference is too heavy ,I use double threshold check to insure image which sended into ML is abnormal.
main code ----> four_classify/monitor.py

PyQt4
numpy
opencv(build with ffmpeg)
matplotlib

relation:
-----
  This is client,  ML is related to 8 layers Neural Networks.