xuebinqin / HRSOD

Towards High-Resolution Salient Object Detection

Introduction

This package contains the source code for Towards High-Resolution Salient Object Detection, ICCV2019.

Datasets

HRSOD [Google Drive] [Baidu Yun](For the link of Baidu Yun, images of HRSOD-Training are splited into two parts just because of single file size limitation.)
DAVIS-S Download

Pre-computed Saliency Maps

Saliency maps of this paper along with compared state-of-the-art methods can be can be downloaded
HRSOD
DAVIS-S

More saliency maps of this paper on widely used benchmarks can be downloaded
THUR
HKU-IS
DUTS-Test
PASCAL-S
DUT-OMRON
ECSSD
HRSOD
DAVIS-S

Source Code

Usage

1. Download our code into your computer

2. Cd to HRSOD-master/caffe-master, follow the official instructions to build caffe. The code has been tested successfully on Ubuntu 14.04 with CUDA 8.0.

3. Make caffe & matcaffe

    make all -j  
    make matcaffe -j  
   

4. Download pretrained caffemodel from baidu yun and put the file under the root directory HRSOD-master/.

5. Change parameters in init_iccv19_demo.m and then run test_iccv19_demo.m to get the saliency maps. The results will be saved in HRSOD-master/results/.

Citation

If you find our works useful in your research, please consider citing:

@InProceedings{Zeng_2019_ICCV,
  author = {Zeng, Yi and Zhang, Pingping and Zhang, Jianming and Lin, Zhe and Lu, Huchuan},
  title = {Towards High-Resolution Salient Object Detection},
  booktitle = {The IEEE International Conference on Computer Vision (ICCV)},
  month = {October},
  year = {2019}
}