Home

Awesome

PANet

Official code repository for paper Parallax-Aware Network for Light Field Salient Object Detection (SPL2024).

Overall Architecture

<p align="center"> <img src="picture/backbone.png" width="100%"/> <br /> <em> Overall architecture of the proposed PANet. </em> </p> <p align="center"> <img src="picture/modules.png" width="80%"/> <br /> <em> Details of the Parallax Processing Module (PPM). </em> </p>

Inference

Download the pretrained model at Baidu Driver (PSW: xf97).

Results

<p align="center"> <img src="picture/results.PNG" width="100%"/> <br /> <em> Quantitative Results. “N/T” denotes unavailable results. ↑/↓ means that a larger/smaller value is better. The top two results are highlighted in <font color="red">red</font> and <font color="blue">blue</font>, respectively. </em> </p>

Download

The saliency maps can be download from Baidu Driver (PSW: kxxi).

Citation

Please cite the following paper if you use this repository in your research:

@article{yuan2024parallax,
  title={Parallax-Aware Network for Light Field Salient Object Detection},
  author={Yuan, Bo and Jiang, Yao and Fu, Keren and Zhao, Qijun},
  journal={IEEE Signal Processing Letters},
  year={2024},
  publisher={IEEE}
}

Any questions regarding this work can contact yuanboera@qq.com.