Hi! I would like to work on your project. Here is as I can see the algorithm:
1. Make some interface to choose the source of the video ( from camera or read from file)
2. Take frame by frame video ( as an image): and here I will work with identification of pedestrians.
3. Change from color to grayscale and use a technique to extract contours( could be with threshold or watershed)
4. Identify actual people, and drop the rest
5. Count and display count bar (amount of people crossing each line)
So I have few question:
1. Are you going to use app for Windows or Linux ?
2. Do you want it in executable or you have a Python IDLE?
3. What the accuracy do you wish to achieve? Last app like this I achieved 97%, if it's ok with you, let's work on it.
My github examples of coding with Python and OpenCv [login to view URL]
Best regards,
Kate