Addressing Class Imbalance in Scene Graph Parsing by Learning to Contrast and Score

He Huang, Shunta Saito, Yuta Kikuchi, Eiichi Matsumoto, Wei Tang, Philip S. Yu

Abstract: Scene graph parsing aims to detect objects in an image scene and recognize their relations. Recent approaches have achieved high av- erage scores on some popular benchmarks, but fail in detecting rare re- lations, as the highly long-tailed distribution of data biases the learning towards frequent labels. Motivated by the fact that detecting these rare relations can be critical in real-world applications, this paper introduces a novel integrated framework of classification and ranking to resolve the class imbalance problem in scene graph parsing. Specifically, we design a new Contrasting Cross-Entropy loss, which promotes the detection of rare relations by suppressing incorrect frequent ones. Furthermore, we propose a novel scoring module, termed as Scorer, which learns to rank the relations based on the image features and relation features to improve the recall of predictions. Our framework is simple and effective, and can be incorporated into current scene graph models. Experimental results show that the proposed approach improves the current state-of-the-art methods, with a clear advantage of detecting rare relations.