FANet: Feature aggregation network for RGBD saliency detection

作者:

Highlights:

• We propose a novel feature aggregation network (FANet) for RGBD saliency detection.

• Region enhanced module REM is used to differentiate salient regions and backgrounds.

• Hierarchical fusion module HFM is used to aggregate multi-modal cues.

• HFM is supported by the graph neural networks (KGNNs) and the non-local module (NLM).

• Extensive experimental results verify the effectiveness of the proposed FANet.

摘要

•We propose a novel feature aggregation network (FANet) for RGBD saliency detection.•Region enhanced module REM is used to differentiate salient regions and backgrounds.•Hierarchical fusion module HFM is used to aggregate multi-modal cues.•HFM is supported by the graph neural networks (KGNNs) and the non-local module (NLM).•Extensive experimental results verify the effectiveness of the proposed FANet.

论文关键词:RGBD saliency,Feature aggregation,Graph neural networks,Hierarchical fusion

论文评审过程:Received 29 June 2020, Revised 7 June 2021, Accepted 8 December 2021, Available online 15 December 2021, Version of Record 3 January 2022.

论文官网地址:https://doi.org/10.1016/j.image.2021.116591