Is Depth Really Necessary for Salient Object Detection?

The PyTorch implementation of the DASNet.

Please see details in http://cvteam.net/projects/2020/DASNet/

Prerequisites

• Python 3.6
• Pytorch 1.0+
• OpenCV 4.0
• Numpy
• TensorboardX
• Apex

DIR

• res: resnet pre-trained models
• eval: test results
• data: datasets

Results

• RGB-D: [NJU2K] [NLPR] [STERE] [DES] [SSD] [LFSD] [SIP]
• RGB: [ECSSD] [DUTS] [DUT-OMRON] [HKU-IS] [PASCAL-S]

Train

RGB-D training sets: RGBD-TR

cd src
python train.py

Test

cd src
python test.py

Evaluation

cd eval
matlab
main

Citation

• If you find this work is helpful, please cite our paper

@inproceedings{zhao2020DASNet,
  title={Is depth really necessary for salient object detection?},
  author={Zhao, Jiawei and Zhao, Yifan and Li, Jia and Chen, Xiaowu},
  booktitle={Proceedings of the 28th ACM International Conference on Multimedia},
  pages={1745--1754},
  year={2020}
}

Reference

This project is based on the following implementations:

• https://github.com/weijun88/F3Net