Adaptive depth-aware visual relationship detection

作者：

Highlights：

• We construct Depth VRDs to make up the lack of depth information of the VRD dataset.

• We propose depth-aware visual fusion module to improve visual appearance features.

• We propose adaptive depth spatial location method to extract depth location.

摘要

•We construct Depth VRDs to make up the lack of depth information of the VRD dataset.•We propose depth-aware visual fusion module to improve visual appearance features.•We propose adaptive depth spatial location method to extract depth location.

论文关键词：Visual relationship detection,Estimated depth maps,Depth-aware visual fusion,Adaptive depth spatial location

论文评审过程：Received 22 November 2021, Revised 16 March 2022, Accepted 8 April 2022, Available online 18 April 2022, Version of Record 28 April 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2022.108786