Learning Human-Human Interactions in Images from Weak Textual Supervision | IEEE Conference Publication | IEEE Xplore