i have a friend who would like a patch that is able to identify all of the red objects in a video and mask everything else in the frame.
i’ve used the basic color-tracking features in jitter before, but i’m not sure how to integrate it to work with multiple objects of a similar color or how to make a mask that conforms to the shape, rather than producing a rectangle around the color.
i know that the jit.cv objects have some things that might be functional for this purpose, but i’m not totally sure how to implement them.
any advice would be greatly appreciated. thanks!
the cv.jit objects will work if you can get the red to become white, that is, if you can get a high contrast between them and the surrounding areas. Might not work very well in a typical scene, cv.jit is great for things like infrared LEDs where there’s super high contrast, but if things are more on a spectrum, there’s lots of thresholding. Just the way it works.
For multiple color tracking you might try the tap.tools demo, I was able to get its tracking working on several blobs of the same/close color, it supports up to 4 (at least when I tried it). Or maybe Cyclops can do this too. Both cost a bit though.
All that said, you may be able to do what you want by honing into the color in question and running a jit.op on it. If the colors are pretty solid, this could work well…if they’re really red, you might be able to use just the red plane, then mask everything below a certain (fairly high) threshold.