Iterative Shrinking for Referring Expression Grounding Using Deep Reinforcement Learning | IEEE Conference Publication | IEEE Xplore