DMCA
Supervised translation-invariant sparse coding (2010)
Cached
Download Links
| Venue: | IN: IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION |
| Citations: | 76 - 8 self |
BibTeX
@INPROCEEDINGS{Yang10supervisedtranslation-invariant,
author = {Jianchao Yang and Kai Yu and Thomas Huang},
title = { Supervised translation-invariant sparse coding},
booktitle = {IN: IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION},
year = {2010},
publisher = {}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
In this paper, we propose a novel supervised hierarchical sparse coding model based on local image descriptors for classification tasks. The supervised dictionary training is performed via back-projection, by minimizing the training error of classifying the image level features, which are extracted by max pooling over the sparse codes within a spatial pyramid. Such a max pooling procedure across multiple spatial scales offer the model translation invariant properties, similar to the Convolutional Neural Network (CNN). Experiments show that our supervised dictionary improves the performance of the proposed model significantly over the unsupervised dictionary, leading to state-of-the-art performance on diverse image databases. Further more, our supervised model targets learning linear features, implying its great potential in handling large scale datasets in real applications.
Keyphrases
translation-invariant sparse coding diverse image database linear feature spatial pyramid large scale datasets state-of-the-art performance classification task hierarchical sparse supervised model target convolutional neural network unsupervised dictionary multiple spatial scale real application local image descriptor sparse code model translation invariant property supervised dictionary training image level feature training error max pooling
