Sparse Coding and Dictionary Learning For Image Analysis: Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro
Sparse Coding and Dictionary Learning For Image Analysis: Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro
Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Optimization for Sparse Coding 1/21
What this part is about
Learning dictionaries with a discriminative cost
function. . .
. . . and a few applications to computer vision
applications.
Compressed sensing with learned dictionaries and
why you should not use random sensing matrices.
Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Optimization for Sparse Coding 2/21
Learning dictionaries with a discriminative cost function
Idea:
Let us consider 2 sets S− , S+ of signals representing 2 different classes.
Each set should admit a specific dictionary best adapted to its
reconstruction.
Classification procedure for a signal x ∈ Rn :
min(R? (x, D− ), R? (x, D+ ))
where
R? (x, D) = minp ||x − Dα||22 s.t. ||α||0 ≤ L.
α∈R
“Reconstructive” training
minD− i∈S− R? (xi , D− )
P
[Grosse et al., 2007], [Huang and Aviyente, 2006] (see also [Wright
et al., 2009])
Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Optimization for Sparse Coding 3/21
Learning dictionaries with a discriminative cost function
“Discriminative” training
[Mairal, Bach, Ponce, Sapiro, and Zisserman, 2008a]
X
min C λzi R? (xi , D− ) − R? (xi , D+ ) ,
D− ,D+
i
Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Optimization for Sparse Coding 4/21
Learning dictionaries with a discriminative cost function
Mixed approach
X
min C λzi R? (xi , D− ) − R? (xi , D+ ) + µR? (xi , Dzi ),
D− ,D+
i
Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Optimization for Sparse Coding 5/21
Learning dictionaries with a discriminative cost function
Examples of dictionaries
Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Optimization for Sparse Coding 7/21
Learning dictionaries with a discriminative cost function
Texture segmentation
Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Optimization for Sparse Coding 8/21
Learning dictionaries with a discriminative cost function
Pixelwise classification
Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Optimization for Sparse Coding 9/21
Learning dictionaries with a discriminative cost function
Multiscale scheme
Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Optimization for Sparse Coding 10/21
Learning dictionaries with a discriminative cost function
weakly-supervised pixel classification
Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Optimization for Sparse Coding 11/21
Application to edge detection and classification
[Mairal, Leordeanu, Bach, Hebert, and Ponce, 2008b]
Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Optimization for Sparse Coding 12/21
Application to edge detection and classification
Berkeley segmentation benchmark
Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Optimization for Sparse Coding 13/21
Application to edge detection and classification
Berkeley segmentation benchmark
Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Optimization for Sparse Coding 14/21
Application to edge detection and classification
Berkeley segmentation benchmark
Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Optimization for Sparse Coding 15/21
Application to edge detection and classification
Contour-based classifier: [Leordeanu, Hebert, and Sukthankar, 2007]
Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Optimization for Sparse Coding 17/21
Application to edge detection and classification
Performance gain due to the prefiltering
Ours + [Leordeanu ’07] [Leordeanu ’07] [Winn ’05]
96.8% 89.4% 76.9%
Recognition rates for the same experiment as [Winn et al., 2005] on
VOC 2005.
Category Ours+[Leordeanu ’07] [Leordeanu ’07]
Aeroplane 71.9% 61.9%
Boat 67.1% 56.4%
Cat 82.6% 53.4%
Cow 68.7% 59.2%
Horse 76.0% 67%
Motorbike 80.6% 73.6%
Sheep 72.9% 58.4%
Tvmonitor 87.7% 83.8%
Average 75.9% 64.2 %
Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Optimization for Sparse Coding 19/21
References I
R. Grosse, R. Raina, H. Kwong, and A. Y. Ng. Shift-invariant sparse coding for audio
classification. In Proceedings of the Twenty-third Conference on Uncertainty in
Artificial Intelligence, 2007.
K. Huang and S. Aviyente. Sparse representation for signal classification. In Advances
in Neural Information Processing Systems, Vancouver, Canada, December 2006.
M. Leordeanu, M. Hebert, and R. Sukthankar. Beyond local appearance: Category
recognition from pairwise interactions of simple features. In Proceedings of the
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2007.
J. Mairal, F. Bach, J. Ponce, G. Sapiro, and A. Zisserman. Discriminative learned
dictionaries for local image analysis. In Proceedings of the IEEE Conference on
Computer Vision and Pattern Recognition (CVPR), 2008a.
J. Mairal, M. Leordeanu, F. Bach, M. Hebert, and J. Ponce. Discriminative sparse
image models for class-specific edge detection and image interpretation. In
Proceedings of the European Conference on Computer Vision (ECCV), 2008b.
J. Winn, A. Criminisi, and T. Minka. Object categorization by learned universal visual
dictionary. In Proceedings of the IEEE International Conference on Computer
Vision (ICCV), 2005.
Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Optimization for Sparse Coding 20/21
References II
J. Wright, A.Y. Yang, A. Ganesh, S.S. Sastry, and Y. Ma. Robust face recognition via
sparse representation. IEEE Transactions on Pattern Analysis and Machine
Intelligence, pages 210–227, 2009.
Francis Bach, Julien Mairal, Jean Ponce and Guillermo Sapiro Optimization for Sparse Coding 21/21