Atom correlation based graph propagation for scene graph generation
作者：
Highlights：
• Atom correlation based graph propagation in the global category space.
• Long-tailed distribution problem alleviation in the scene graph generation task.
• Performance superiority over state-of-the-art methods and strong baselines.
• Model capability of capturing infrequent and missed relationships.
• Flexible and holistic introduction of human commonsense knowledge.
摘要
•Atom correlation based graph propagation in the global category space.•Long-tailed distribution problem alleviation in the scene graph generation task.•Performance superiority over state-of-the-art methods and strong baselines.•Model capability of capturing infrequent and missed relationships.•Flexible and holistic introduction of human commonsense knowledge.
论文关键词：Scene graph generation,Long-tailed distribution,Knowledge graph,Atom correlation,Category space
论文评审过程：Received 25 July 2020, Revised 11 May 2021, Accepted 31 August 2021, Available online 3 September 2021, Version of Record 9 September 2021.
论文官网地址：https://doi.org/10.1016/j.patcog.2021.108300
