Classification with invariant scattering representations

Joan Bruna; Stéphane Mallat

doi:10.1109/IVMSPW.2011.5970362

Classification with invariant scattering representations

Joan Bruna, Stéphane Mallat

Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

A scattering transform defines a signal representation which is invariant to translations and Lipschitz continuous relatively to deformations. It is implemented with a non-linear convolution network that iterates over wavelet and modulus operators. Lipschitz continuity locally linearizes deformations. Complex classes of signals and textures can be modeled with low-dimensional affine spaces, computed with a PCA in the scattering domain. Classification is performed with a penalized model selection. State of the art results are obtained for handwritten digit recognition over small training sets, and for texture classification.

Original language	English (US)
Title of host publication	2011 IEEE 10th IVMSP Workshop
Subtitle of host publication	Perception and Visual Signal Analysis, IVMSP 2011 - Proceedings
Publisher	IEEE Computer Society
Pages	99-104
Number of pages	6
ISBN (Print)	9781457712869
DOIs	https://doi.org/10.1109/IVMSPW.2011.5970362
State	Published - 2011

Publication series

Name	2011 IEEE 10th IVMSP Workshop: Perception and Visual Signal Analysis, IVMSP 2011 - Proceedings

Keywords

Image classification
Invariant representations
local image descriptors
pattern recognition
texture classification

ASJC Scopus subject areas

Computer Vision and Pattern Recognition
Signal Processing

Access to Document

10.1109/IVMSPW.2011.5970362

Cite this

Bruna, J., & Mallat, S. (2011). Classification with invariant scattering representations. In 2011 IEEE 10th IVMSP Workshop: Perception and Visual Signal Analysis, IVMSP 2011 - Proceedings (pp. 99-104). Article 5970362 (2011 IEEE 10th IVMSP Workshop: Perception and Visual Signal Analysis, IVMSP 2011 - Proceedings). IEEE Computer Society. https://doi.org/10.1109/IVMSPW.2011.5970362

Classification with invariant scattering representations. / Bruna, Joan; Mallat, Stéphane.
2011 IEEE 10th IVMSP Workshop: Perception and Visual Signal Analysis, IVMSP 2011 - Proceedings. IEEE Computer Society, 2011. p. 99-104 5970362 (2011 IEEE 10th IVMSP Workshop: Perception and Visual Signal Analysis, IVMSP 2011 - Proceedings).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

@inproceedings{3304a7becbfb4fc991a389896866a002,

title = "Classification with invariant scattering representations",

abstract = "A scattering transform defines a signal representation which is invariant to translations and Lipschitz continuous relatively to deformations. It is implemented with a non-linear convolution network that iterates over wavelet and modulus operators. Lipschitz continuity locally linearizes deformations. Complex classes of signals and textures can be modeled with low-dimensional affine spaces, computed with a PCA in the scattering domain. Classification is performed with a penalized model selection. State of the art results are obtained for handwritten digit recognition over small training sets, and for texture classification.",

keywords = "Image classification, Invariant representations, local image descriptors, pattern recognition, texture classification",

author = "Joan Bruna and St{\'e}phane Mallat",

year = "2011",

doi = "10.1109/IVMSPW.2011.5970362",

language = "English (US)",

isbn = "9781457712869",

series = "2011 IEEE 10th IVMSP Workshop: Perception and Visual Signal Analysis, IVMSP 2011 - Proceedings",

publisher = "IEEE Computer Society",

pages = "99--104",

booktitle = "2011 IEEE 10th IVMSP Workshop",

}

TY - GEN

T1 - Classification with invariant scattering representations

AU - Bruna, Joan

AU - Mallat, Stéphane

PY - 2011

Y1 - 2011

N2 - A scattering transform defines a signal representation which is invariant to translations and Lipschitz continuous relatively to deformations. It is implemented with a non-linear convolution network that iterates over wavelet and modulus operators. Lipschitz continuity locally linearizes deformations. Complex classes of signals and textures can be modeled with low-dimensional affine spaces, computed with a PCA in the scattering domain. Classification is performed with a penalized model selection. State of the art results are obtained for handwritten digit recognition over small training sets, and for texture classification.

AB - A scattering transform defines a signal representation which is invariant to translations and Lipschitz continuous relatively to deformations. It is implemented with a non-linear convolution network that iterates over wavelet and modulus operators. Lipschitz continuity locally linearizes deformations. Complex classes of signals and textures can be modeled with low-dimensional affine spaces, computed with a PCA in the scattering domain. Classification is performed with a penalized model selection. State of the art results are obtained for handwritten digit recognition over small training sets, and for texture classification.

KW - Image classification

KW - Invariant representations

KW - local image descriptors

KW - pattern recognition

KW - texture classification

UR - http://www.scopus.com/inward/record.url?scp=80052302973&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=80052302973&partnerID=8YFLogxK

U2 - 10.1109/IVMSPW.2011.5970362

DO - 10.1109/IVMSPW.2011.5970362

M3 - Conference contribution

AN - SCOPUS:80052302973

SN - 9781457712869

T3 - 2011 IEEE 10th IVMSP Workshop: Perception and Visual Signal Analysis, IVMSP 2011 - Proceedings

SP - 99

EP - 104

BT - 2011 IEEE 10th IVMSP Workshop

PB - IEEE Computer Society

ER -

Classification with invariant scattering representations

Abstract

Publication series

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this