StimCollection class about pliers HOT 3 CLOSED

This would be nice. Also, if it had multiprocessing, so this could be easily parallelized. A related question is about how we want to deal with features that could be applied to video and image stims. Do we want to loop over all frames and treat them as images?

from pliers.

Re: images vs videos: that's how it works right now, yes. You can iterate VideoStims, which gives you VideoFrameStims that are basically glorified images. This works okay, but definitely needs to be made more principled. Right now every extractor that can handle both needs to explicitly check for type and adjust accordingly. It's probably worth spending some time to come up with a cleaner interface (or at the very least, a helpful Mixin) that minimizes redundancy and imposes some standardization.

from pliers.

Closing because we're moving to a scikit-learn-like API that expects most extraction calls to happen on the Extractor rather than the Stim (i.e., extractor.transform() rather than stim.extract()).

from pliers.