G. Hinton, A. Krizhevsky, and S. Wang. Artificial Neural Networks and Machine Learning -- ICANN 2011, page 44--51. Berlin, Heidelberg, Springer Berlin Heidelberg, (2011)
Abstract
The artificial neural networks that are used to recognize shapes typically use one or more layers of learned feature detectors that produce scalar outputs. By contrast, the computer vision community uses complicated, hand-engineered features, like SIFT 6, that produce a whole vector of outputs including an explicit representation of the pose of the feature. We show how neural networks can be used to learn features that output a whole vector of instantiation parameters and we argue that this is a much more promising way of dealing with variations in position, orientation, scale and lighting than the methods currently employed in the neural networks community. It is also more promising than the hand-engineered features currently used in computer vision because it provides an efficient way of adapting the features to the domain.
%0 Conference Paper
%1 10.1007/978-3-642-21735-7_6
%A Hinton, Geoffrey E.
%A Krizhevsky, Alex
%A Wang, Sida D.
%B Artificial Neural Networks and Machine Learning -- ICANN 2011
%C Berlin, Heidelberg
%D 2011
%E Honkela, Timo
%E Duch, Włodzisław
%E Girolami, Mark
%E Kaski, Samuel
%I Springer Berlin Heidelberg
%K readinggroup
%P 44--51
%T Transforming Auto-Encoders
%X The artificial neural networks that are used to recognize shapes typically use one or more layers of learned feature detectors that produce scalar outputs. By contrast, the computer vision community uses complicated, hand-engineered features, like SIFT 6, that produce a whole vector of outputs including an explicit representation of the pose of the feature. We show how neural networks can be used to learn features that output a whole vector of instantiation parameters and we argue that this is a much more promising way of dealing with variations in position, orientation, scale and lighting than the methods currently employed in the neural networks community. It is also more promising than the hand-engineered features currently used in computer vision because it provides an efficient way of adapting the features to the domain.
%@ 978-3-642-21735-7
@inproceedings{10.1007/978-3-642-21735-7_6,
abstract = {The artificial neural networks that are used to recognize shapes typically use one or more layers of learned feature detectors that produce scalar outputs. By contrast, the computer vision community uses complicated, hand-engineered features, like SIFT [6], that produce a whole vector of outputs including an explicit representation of the pose of the feature. We show how neural networks can be used to learn features that output a whole vector of instantiation parameters and we argue that this is a much more promising way of dealing with variations in position, orientation, scale and lighting than the methods currently employed in the neural networks community. It is also more promising than the hand-engineered features currently used in computer vision because it provides an efficient way of adapting the features to the domain.},
added-at = {2020-03-24T14:44:20.000+0100},
address = {Berlin, Heidelberg},
author = {Hinton, Geoffrey E. and Krizhevsky, Alex and Wang, Sida D.},
biburl = {https://www.bibsonomy.org/bibtex/2784b8ebeb6a8146479b8d67448a74eb7/lautenschlager},
booktitle = {Artificial Neural Networks and Machine Learning -- ICANN 2011},
description = {Transforming Auto-Encoders | SpringerLink},
editor = {Honkela, Timo and Duch, W{\l}odzis{\l}aw and Girolami, Mark and Kaski, Samuel},
interhash = {65220918699bf6d3dd721d22d44abd01},
intrahash = {784b8ebeb6a8146479b8d67448a74eb7},
isbn = {978-3-642-21735-7},
keywords = {readinggroup},
pages = {44--51},
publisher = {Springer Berlin Heidelberg},
timestamp = {2020-03-24T14:44:20.000+0100},
title = {Transforming Auto-Encoders},
year = 2011
}