Awesome
IIAU (2017) Saliency Detection Models
Introduction
This file contains our saliency detection models appeared in ICCV2017. We delve into the convolutional features in pre-trained FCN models and propose the following methods.
- Amulet
Title: Amulet: Aggregating Multi-level Convolutional Features for Salient Object Detection
Authors: Pingping Zhang, Dong Wang, Huchuan Lu*, Hongyu Wang and Xiang Ruan
Paper Links:
Google Drive Link: https://drive.google.com/file/d/0B5t8yubOMmTKUnZUYTY5NDdWcTg/view?usp=sharing
One Drive Link: https://1drv.ms/u/s!AnzjJimFPB2Ibqy-4-sxZOHnb2s
Baidu Pan Link: https://pan.baidu.com/s/1gfKkIwN
- UCF
Title: Learning Uncertain Convolutional Features for Accurate Saliency Detection
Authors: Pingping Zhang, Dong Wang, Huchuan Lu*, Hongyu Wang and Baocai Yin
Paper Links:
Google Drive Link: https://drive.google.com/file/d/0B5t8yubOMmTKbUFUQ1gtekVnNEE/view?usp=sharing
One Drive Link: https://1drv.ms/u/s!AnzjJimFPB2IbY_bYGa-d5HemFw
Baidu Pan Link: https://pan.baidu.com/s/1skCi4Dv
- SRM
Title: A Stagewise Refinement Model for Detecting Salient Objects in Images
Authors: Tiantian Wang, Ali Borji,Lihe Zhang,Pingping Zhang, Huchuan Lu*
Paper Links:
Google Drive Link: https://drive.google.com/file/d/0B_MpGgTntG47b18yUlJxOGtuNDA/view
One Drive Link: https://1drv.ms/u/s!AnzjJimFPB2IbY_bYGa-d5HemFw
Baidu Pan Link: https://pan.baidu.com/s/1skCi4Dv
How to use
Prerequisites:
Download source code from
Google Drive Link: https://drive.google.com/open?id=0B5t8yubOMmTKazhHdV8yR19tek0
One Drive Link: https://1drv.ms/u/s!AnzjJimFPB2Ib_zU3E15C6AxkXQ
Baidu Pan Link: http://pan.baidu.com/s/1bTAYNc
Follow the official websites of the Caffe framework and install the whole toolbox (necessary Matlab wrappers)
- Training
The training code is in ./models/Amulet and ./models/UCF. For saving memory, we use the BN method implemented by Alex Kendall etal http://mi.eng.cam.ac.uk/projects/segnet/. The training is followed as examples in Caffe. We use the ImageData layer to load training image pairs. The data list can be generated by Matlab in
Google Drive Link: https://drive.google.com/open?id=0B5t8yubOMmTKblFZajBuTTYyanc
Baidu Pan Link: http://pan.baidu.com/s/1pKXnBdD
The only need is changing the path of data files.
- Testing
The testing code is in ./matlab/Amulet_test and ./matlab/UCF_test.
For saliency testing,
(1) Get prediction : test_saliency_dataset.m
(2) Get PR value : PR.m
(3) Plot PR curves: code_pr.m
(4) Plot Bar figures: code_bar.m
(5) Get MAE for each method : getmae.m
Note that we have provided the PR curves in ./PR_curves
Results
The saliency maps can be download via baidu yun at http://pan.baidu.com/s/1o7O3lQq.
Citation
Please cite the following papers if our models help your research:
@article{zhang2017amulet,
title={Amulet: Aggregating Multi-level Convolutional Features for Salient Object Detection},
author={Zhang, Pingping and Wang, Dong and Lu, Huchuan and Wang, Hongyu and Ruan, Xiang},
journal={arXiv preprint arXiv:1708.02001},
year={2017}
}
@InProceedings{Zhang_2017_ICCV,
author = {Zhang, Pingping and Wang, Dong and Lu, Huchuan and Wang, Hongyu and Ruan, Xiang},
title = {Amulet: Aggregating Multi-Level Convolutional Features for Salient Object Detection},
booktitle = {The IEEE International Conference on Computer Vision (ICCV)},
month = {Oct},
year = {2017}
}
@article{zhang2017learning,
title={Learning Uncertain Convolutional Features for Accurate Saliency Detection},
author={Zhang, Pingping and Wang, Dong and Lu, Huchuan and Wang, Hongyu and Yin, Baocai},
journal={arXiv preprint arXiv:1708.02031},
year={2017}
}
@InProceedings{Zhang_2017_ICCV,
author = {Zhang, Pingping and Wang, Dong and Lu, Huchuan and Wang, Hongyu and Yin, Baocai},
title = {Learning Uncertain Convolutional Features for Accurate Saliency Detection},
booktitle = {The IEEE International Conference on Computer Vision (ICCV)},
month = {Oct},
year = {2017}
}
@InProceedings{Wang_2017_ICCV,
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)},
month = {Oct},
year = {2017}
}
Question and connection
If any question, please connect