Awesome

SAC-Net: Spatial Attenuation Context for Salient Object Detection

by Xiaowei Hu, Chi-Wing Fu, Lei Zhu, Tianyu Wang, and Pheng-Ann Heng.

This implementation is written by Xiaowei Hu at the Chinese University of Hong Kong.

Citation

@article{hu2021sac,
  title={{SAC-Net}: Spatial attenuation context for salient object detection},
  author={Hu, Xiaowei and Fu, Chi-Wing and Zhu, Lei and Wang, Tianyu and Heng, Pheng-Ann},
  journal={IEEE Transactions on Circuits and Systems for Video Technology},
  volume={31},
  number={3},
  pages={1079--1090},
  year={2021}
}

Saliency Maps

Please find the predicted saliency maps on the ECSSD, PASCAL-S, SOD, HKU-IS, DUT-OMRON, and DUTS-test at Google Drive.

Installation

Please download and compile our CF-Caffe.
Put the examples/SAC-Net/ into CF-Caffe/examples/.

Train

Download the ResNet-101 or ResNet-50 model trained on the ImageNet and put this model in CF-Caffe/models/.

Enter the ./examples/SAC-Net/SAC-Net-res101/ or ./examples/SAC-Net/SAC-Net-res50/. Modify the image path in train_val.prototxt.
Run
```
sh train.sh
```

Test

Put the trained model in ./examples/SAC-Net/SAC-Net-res101/snapshot/.
Enter the ./examples/ and run test_saliency.m in Matlab.