2 ms·
I've done a few computer vision projects lately and it's really amazing what you can get done with opencv for relatively clean data. In this case you're trying
by dekhn 8y ago
I've done a few computer vision projects lately and it's really amazing what you can get done with opencv for relatively clean data. In this case you're trying to identify circles and classify their color into one of several groups. This is pretty common, the main challenge is cleaning up the raw data so that the hough transform runs quickly and accurately, and the color cluster identifier is robust.
If you're pretty good with Tensorflow, you could do it with an object detector. I've built absurdly accurate object detectors using the TF Object Detector tutorial and just a little hand labelled data plus some good synthetic augmentation.
For this project's scope and scale it's probably not worth automating unless you're going to be repeatedly running the process.