GitHub - huchuanlu/13

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 1 Commit
Bayes.m		Bayes.m
PCA.m		PCA.m
RGB2Lab.m		RGB2Lab.m
ReadDAT.m		ReadDAT.m
SLICSuperpixelSegmentation.exe		SLICSuperpixelSegmentation.exe
calOptimizedGuassTemplate.m		calOptimizedGuassTemplate.m
calculateBayesPosterior.m		calculateBayesPosterior.m
calculateGuassOptimization.m		calculateGuassOptimization.m
calculatePCAError.m		calculatePCAError.m
calculateRegionProps.m		calculateRegionProps.m
calculateSRError.m		calculateSRError.m
chooseFeature.m		chooseFeature.m
combineSalMap.m		combineSalMap.m
convertRecErrorToSal.m		convertRecErrorToSal.m
deletebd.m		deletebd.m
demo.m		demo.m
descendPropagation.m		descendPropagation.m
extractSupfeat.m		extractSupfeat.m
extract_bg_sp.m		extract_bg_sp.m
form_codebook.m		form_codebook.m
getSlicLabelMap.m		getSlicLabelMap.m
labelCluster.m		labelCluster.m
mexLasso.m		mexLasso.m
mexLasso.mexw32		mexLasso.mexw32
mexLasso.mexw64		mexLasso.mexw64
normVector.m		normVector.m
readme.txt		readme.txt
test1.bmp		test1.bmp
test1.dat		test1.dat
test1_SLIC.bmp		test1_SLIC.bmp
test2.bmp		test2.bmp
test2.dat		test2.dat
test2_SLIC.bmp		test2_SLIC.bmp
test3.bmp		test3.bmp
test3.dat		test3.dat
test3_SLIC.bmp		test3_SLIC.bmp
vgg_argparse.m		vgg_argparse.m
vgg_kmeans.m		vgg_kmeans.m
vgg_kmiter.cxx		vgg_kmiter.cxx
vgg_kmiter.dll		vgg_kmiter.dll
vgg_kmiter.mexglx		vgg_kmiter.mexglx
vgg_kmiter.mexw64		vgg_kmiter.mexw64

Repository files navigation

The source code implements the saliency detection algorithm described in the published work [1].
This code is the preliminary version. We appreciate any comments/suggestions.
Questions regarding the code can be directed to Xiaohui Li at [email protected].

*********************************************************************************************************************************************************************************

Before you use the code, please make sure that mex files are correctly compiled.
Some functions like mexLasso.m need the libararies in the SPAMS toolbox:
https://www.di.ens.fr/willow/SPAMS/downloads.html

*********************************************************************************************************************************************************************************

The code runs on Windows XP with MATLAB R2009b.

To get a quick overview:
1. Add your test image in the current directory.
2. Edit the image name 'imName' in demo.m.
3. Run demo.m.

You can also set/change the parameters in demo.m if necessary.

After running demo.m, you can obtain seven saliency maps of different phases as presented in [1] and the final integrated saliency map is named as '****_DSR.bmp'.

*********************************************************************************************************************************************************************************

Note:
1. We observe that some images on the MSRA dataset are surrounded with artificial frames, which will invalidate the used boundary templates. Therefore, we run a pre-processing to remove such obvious frames. If necessary, you can refer to deletebd.m for more details.
2. We utilize the SLIC execution file 'SLICSuperpixelSegmentation.exe' of the published work [2] ( https://ivrg.epfl.ch/research/superpixels ). Make sure the input image be .bmp file to execute 'SLICSuperpixelSegmentation.exe'.

*********************************************************************************************************************************************************************************

References:
[1] Xiaohui Li, Huchuan Lu, Ming-Hsuan Yang, Lihe Zhang and Xiang Ruan. Saliency Detection Via Dense and Sparse Reconstruction. International Conference on Computer Vision (ICCV), 2013.
[2] R. Achanta, A. Shaji, K. Smith, A. Lucchi, P. Fua, and S. Susstrunk. Slic superpixels. Technical Report, EPFL, 2010.