Collaborative Deep Reinforcement Learning for Joint Object Search

摘要

We examine the problem of joint top-down active search of multiple objectsunder interaction, e.g., person riding a bicycle, cups held by the table, etc..Such objects under interaction often can provide contextual cues to each otherto facilitate more efficient search. By treating each detector as an agent, wepresent the first collaborative multi-agent deep reinforcement learningalgorithm to learn the optimal policy for joint active object localization,which effectively exploits such beneficial contextual information. We learninter-agent communication through cross connections with gates between theQ-networks, which is facilitated by a novel multi-agent deep Q-learningalgorithm with joint exploitation sampling. We verify our proposed method onmultiple object detection benchmarks. Not only does our model help to improvethe performance of state-of-the-art active localization models, it also revealsinteresting co-detection patterns that are intuitively interpretable.