forked from Pchank/caffe-sal
-
Notifications
You must be signed in to change notification settings - Fork 0
/
IIAU_saliency2018
113 lines (61 loc) · 3.22 KB
/
IIAU_saliency2018
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
IIAU (2018) Saliency Detection Models
Introduction
------------------------------------------------------------------------------------------------------------------
This file contains our saliency detection models appeared in 2018.
1) LFR (IJCAI2018)
Title: Salient Object Detection by Lossless Feature Reflection
Authors: Pingping Zhang, Wei Liu, Huchuan Lu* and Chunhua Shen
Paper Links: https://arxiv.org/abs/1802.06527
2) HyperFusion-Net
Title: HyperFusion-Net: Densely Reflective Fusion for Salient Object Detection
Authors: Pingping Zhang, Huchuan Lu* and Chunhua Shen
Paper Links: https://arxiv.org/abs/1804.05142
3) Agile Amulet
Title: Agile Amulet: Real-Time Salient Object Detection with Contextual Attention
Authors: Pingping Zhang, Luyao Wang, Dong Wang, Huchuan Lu* and Chunhua Shen
Paper Links: https://arxiv.org/abs/1804.05142
How to use
--------------------------------------------------------------------------------------------------------------
Prerequisites:
Download source code from Link: https://1drv.ms/u/s!AnzjJimFPB2IcMN1agL18fD18Us
Follow the official websites of the Caffe framework and install the whole toolbox (necessary Matlab wrappers)
1) Training
The training code is in ./models/T-Net,./models/HDFN and ./models/AAmulet. 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.
2) Testing
The testing code is in ./matlab/.
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
Citation
---------------------------------------------------------------------------------------------------------------------
Please cite the following papers if our models help your research:
@article{zhang2018salient,
title={Salient Object Detection by Lossless Feature Reflection},
author={Zhang, Pingping and Liu, Wei and Lu, Huchuan and Shen, Chunhua},
journal={arXiv preprint arXiv:1802.06527},
year={2018}
}
@article{zhang2018hyperfusion,
title={HyperFusion-Net: Densely Reflective Fusion for Salient Object Detection},
author={Zhang, Pingping and Lu, Huchuan and Shen, Chunhua},
journal={arXiv preprint arXiv:1804.05142},
year={2018}
}
@article{zhang2018agile,
title={Agile Amulet: Real-Time Salient Object Detection with Contextual Attention},
author={Zhang, Pingping and Wang, Luyao and Wang, Dong and Lu, Huchuan and Shen, Chunhua},
journal={arXiv preprint arXiv:1802.06960},
year={2018}
}
Question and connection
------------------------------------------------------------------------------------------------------------------
If any question, please connect
jssxzhpp@gmail.com; jssxzhpp@mail.dlut.edu.cn