Rethinking the U-shape Structure for Salient Object Detection

This is the official PyTorch implementation of our TIP 2021 paper.

Prerequisites

• Pytorch 0.4.1+
• torchvision

Usage

1. Clone the repository

git clone https://github.com/zal0302/CII.git
cd CII/

2. Download the datasets

Download the following datasets for testing and unzip them into data folder.

3. Download the pre-trained models for CII and backbone

Download the following pre-trained models for CII with ResNet50 backbone and ResNet18 backbone into saved/models folder.

4. Test

For all datasets testing used in our paper for ResNet50 backbone:

python test.py -r saved/models/cii.pth -c saved/models/config.json

and for ResNet18 backbone:

python test.py -r saved/models/cii_res18.pth -c saved/models/config_resnet18.json

All results saliency maps will be stored under saved/results folders in .png formats.

5. Pre-computed results and evaluation results

You may refer to this repo for results evaluation: SalMetric.

We provide the pre-computed saliency maps and evaluation results for ResNet50 backbone and ResNet18 backbone.

6. Contact

If you have any questions, feel free to contact me via: liuzhiang(at)mail.nankai.edu.cn.

If you think this work is helpful, please cite

@article{liu2021rethinking,
  title={Rethinking the U-Shape Structure for Salient Object Detection},
  author={Liu, Jiang-Jiang and Liu, Zhi-Ang and Peng, Pai and Cheng, Ming-Ming},
  journal={IEEE Transactions on Image Processing},
  volume={30},
  pages={9030--9042},
  year={2021},
  publisher={IEEE}
}

@article{liu2022poolnet+,
  title={Poolnet+: Exploring the potential of pooling for salient object detection},
  author={Liu, Jiang-Jiang and Hou, Qibin and Liu, Zhi-Ang and Cheng, Ming-Ming},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year={2022},
  publisher={IEEE}
}