CV in openFrameworks


#1

Following up on a recent conversation with @bakercp and @mikewesthad - what are the best tools to do CV image analysis on video for openFrameworks?

I’m particularly interested in being able to pull out moving bodies from a black and white film. Should I stick with the core ofxOpenCV addon? Have either of you tried @kylemcdonald’s ofxCv? Would you recommend any of the many blob detection kits in ofxAddons?

Any other useful CV related reference material is welcome. Thanks!


#2

ofxCv is great. But before you go down that rabbit hole I’d recommend defining your “problem” a bit more. Find some examples of what you’d like to accomplish, etc. Rotoscoping? Background subtraction? Foreground extraction?

Maybe it’s better to do the rotoscoping using an assistive tool like After Effects and save a bunch of masks that you could apply (and use no computer vision)?

Anyway … computer vision is awesome and often promises way more than it can deliver, at least initially …


#3

Thanks for the prompt and as usual totally on point response, @bakercp! I haven’t used after effects ever, but after watching a short tutorial on auto rotoscoping, it seems like that is probably the way to go. But will definitely do some more research and think more about exactly what the problem I’m trying to solve is :smile:


#4

Yeah CV always looks cool but it can get pretty complex and the abstraction involved with a lot of the addons can make things frustrating when it doesn’t quite do what you want. Since you are working in a static medium after effects might be easier in the long run unless you are doing some sort of dynamic manipulation to the films. It will take more time probably but your results will ultimately be more stable and capture is way easier since you just export the new movie aka easier to document and nicer in a portfolio.