Extracted features from a vision model. I haven't experimented with CLIP yet but would like to as I think adding clothing search would be interesting