# Learning Synergistic Attention for Light Field Salient Object Detection

**Authors:** Yi Zhang, Geng Chen, Qian Chen, Yujia Sun, Yong Xia, Olivier Deforges, Wassim Hamidouche, Lu Zhang  
**arXiv:** [2104.13916](https://arxiv.org/abs/2104.13916)  
**Date:** 2021-04-28  
**Summary page:** https://www.summarizepaper.com/en/paper/2104.13916v4/

## Key points

- Yi Zhang and colleagues propose a novel Synergistic Attention Network (SA-Net) for light field salient object detection.
- SA-Net establishes a synergistic effect between multi-modal features with advanced attention mechanisms, exploiting the rich information of focal stacks via 3D convolutional neural networks.
- The SA-Net decodes the high-level features of multi-modal light field data with two cascaded synergistic attention modules and predicts the saliency map using an effective feature fusion module in a progressive manner.
- Extensive experiments on three widely-used benchmark datasets demonstrate that SA-Net outperforms 28 state-of-the-art models, sufficiently demonstrating its effectiveness and superiority.
- The proposed method is expected to have significant implications for various applications such as autonomous driving, robotics, and virtual reality.
- The code for SA-Net is available at https://github.com/PanoAsh/SA-Net which allows further exploration of this promising approach to tackle the challenging problem of light field salient object detection by leveraging synergistic attention mechanisms and multi-modal features.

## Summary

In their paper titled "Learning Synergistic Attention for Light Field Salient Object Detection," Yi Zhang and colleagues propose a novel Synergistic Attention Network (SA-Net) to address the challenge of light field salient object detection. The proposed SA-Net establishes a synergistic effect between multi-modal features with advanced attention mechanisms, exploiting the rich information of focal stacks via 3D convolutional neural networks. The SA-Net decodes the high-level features of multi-modal light field data with two cascaded synergistic attention modules and predicts the saliency map using an effective feature fusion module in a progressive manner. The authors conducted extensive experiments on three widely-used benchmark datasets, demonstrating that their SA-Net outperforms 28 state-of-the-art models, sufficiently demonstrating its effectiveness and superiority. The proposed method is expected to have significant implications for various applications such as autonomous driving, robotics, and virtual reality. The code for SA-Net is available at https://github.com/PanoAsh/SA-Net which allows further exploration of this promising approach to tackle the challenging problem of light field salient object detection by leveraging synergistic attention mechanisms and multi-modal features.