DMCA
Harvesting mid-level visual concepts from large-scale internet images (2013)
| Venue: | In CVPR |
| Citations: | 19 - 4 self |
BibTeX
@INPROCEEDINGS{Li13harvestingmid-level,
author = {Quannan Li and Jiajun Wu and Zhuowen Tu},
title = {Harvesting mid-level visual concepts from large-scale internet images},
booktitle = {In CVPR},
year = {2013}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Obtaining effective mid-level representations has be-come an increasingly important task in computer vision. In this paper, we propose a fully automatic algorithm which harvests visual concepts from a large number of Internet images (more than a quarter of a million) using text-based queries. Existing approaches to visual concept learning from Internet images either rely on strong supervision with detailed manual annotations or learn image-level classifiers only. Here, we take the advantage of having massive well-organized Google and Bing image data; visual concepts (around 14, 000) are automatically exploited from images using word-based queries. Using the learned visual con-cepts, we show state-of-the-art performances on a variety of benchmark datasets, which demonstrate the effectiveness of the learned mid-level representations: being able to gen-eralize well to general natural images. Our method shows significant improvement over the competing systems in im-age classification, including those with strong supervision. 1.
Keyphrases
mid-level visual concept large-scale internet image visual concept internet image strong supervision automatic algorithm massive well-organized google detailed manual annotation large number word-based query learned visual con-cepts significant improvement computer vision benchmark datasets general natural image important task image-level classifier im-age classification learned mid-level representation effective mid-level representation text-based query state-of-the-art performance bing image data
