A unified deep sparse graph attention network for scene graph generation

作者：

Highlights：

• Propose an efficient framework of feature interaction and knowledge learning for SGG.

• Construct sparse graph by classifying dense edges into foreground and background.

• Jointly learn object and relationship features via graphical message passing in GAT.

• Explore the prior information of statistical probability and build sparse knowledge graph.

• Our method outperforms several state-of-the-art methods in VG datasets.

摘要

•Propose an efficient framework of feature interaction and knowledge learning for SGG.•Construct sparse graph by classifying dense edges into foreground and background.•Jointly learn object and relationship features via graphical message passing in GAT.•Explore the prior information of statistical probability and build sparse knowledge graph.•Our method outperforms several state-of-the-art methods in VG datasets.

论文关键词：Scene graph generation,Statistical co-occurrence knowledge,Relationship measurement network,Graph attention network,Sparse graph

论文评审过程：Received 18 February 2021, Revised 30 June 2021, Accepted 6 October 2021, Available online 8 October 2021, Version of Record 16 October 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108367