PMP-NET: RETHINKING VISUAL CONTEXT FOR SCENE GRAPH GENERATION [PDF]
Published in ICASSP, 2022
Scene graph generation aims to describe the contents of scenes by identifying the objects and their relationships. In previous works, visual context is widely utilized in message-passing networks to generate the representations for classification. However, the noisy estimation of visual context limits model performance. In this paper, we revisit the visual context propagation in SGG and propose a Progressive Message Passing Network to estimate context in a coarse-to-fine manner. 