Search terms for : object detection in video, bounds crossing on video

asked 2018-10-04 06:02:55 -0500

russelldb gravatar image

Hi, I have a real would application I'm developing, as a complete beginner to openvc. I'm using python3, and I've bought the pyimagesearch book, which I am working through (and enjoying.) However, I'm starting to think about what search terms I need to know in order to look for the solutions to my real world problem. I'm so ignorant I don't even know how to look for answers yet!

The use case: detecting laps on a velodrome. Given a static camera pointing at a section of track, I'd like to declare a space in the frame as being the finish line, and detect whenever a bike crosses the line. For this I think there are quite a few examples online, but what techniques should I be searching for to learn?

The harder part (is it?) Detecting WHO crossed the line. Given only four riders, and I can stick a patch that is unique and known in advance (like a template) on each rider's helmet, or shoe, or whatever. How can I detect the patches in live video? Should I be searching for 'template detection'? 'object detection?' 'patch detection'? Again, what are the techniques I need?

BTW if you think you can do this use case, and freelance, contact me.

Cheers

Russell

edit retag flag offensive close merge delete

Comments

google for cnn - convolutionary neuronal network - read a bit about the basics and then pick a framework. (tensorflow, darknet yolo,torch,caffe)

For object detection i personally like to use yolo.

holger gravatar imageholger ( 2018-10-05 00:02:25 -0500 )edit