My goal is to recognize specific types of traffic signs: red circles on video in real time.
Scikit-Image and OpenCV are both great for their respective uses. Unfortunately, as far as I know, Scikit-Image doesn't have the ability to do real-time video. OpenCV is much better for this problem.
If you don't know where to begin, here's a great tutorial series. It is exactly what you're asking, just substitute a stop-sign trained classifier for the face trained classifier. https://pythonprogramming.net/haar-cascade-face-eye-detection-python-opencv-tutorial/
Use a Haar cascade for the sign detection. I found with a quick search for 'Stop sign Haar cascade' that other people already have created trained classifiers like this http://www.cs.utah.edu/~turcsans/DUC_files/HaarCascades/
On the page, download haarcascade_stop.xml