A Stagewise Refinement Model for Detecting Salient Objects in Images (SRM)

This package has the source code for the paper "A Stagewise Refinement Model for Detecting Salient Objects in Images" (ICCV17).

Paper link

• The paper can be found in Baidu drive or Google drive.

Note

• We use Caffe to train our Stagewise Refinement Model.
• Our Resnet-50 model is based on the previous work simon2016cnnmodels "Imagenet pretrained models with batch normalization".
• Saliency maps are generated without any pre- or post-processing.

How to use

Train

• Download our initialized model from Baidu drive or Google drive.
• Use the code in ./train to train the network.

Test

• Download our trained model from Baidu drive or Google drive.
• Run ./test/test.m to generate saliency maps in the ./saliency_map folder.

Download

• The saliency maps on 10 datasets including ECSSD, PASCAL-S, SOD, SED1, SED2, MSRA, DUT-OMRON, THUR15K, HKU-IS and DUTS can be found in Baidu drive or Google drive.

Cite this work

If you find this work useful in your research, please consider citing:

 @inproceedings{wangiccv17,
    author={Wang, Tiantian and Borji, Ali and Zhang, Lihe and Zhang, Pingping and Lu, Huchuan},
    title={A Stagewise Refinement Model for Detecting Salient Objects in Images},
    booktitle = {The IEEE International Conference on Computer Vision (ICCV)},
    year = {2017}
 }

Contact

[email protected]