Comparative Object Similarity for Improved Recognition with Few or No Examples
Abstract
Learning models for recognizing objects with few or no training examples is important, due to the intrinsic long-tailed distribution of objects in the real world. In this paper, we propose an approach to use comparative object similarity. The key insight is that: given a set of object categories which are similar and a set of categories which are dissimilar, a good object model should respond more strongly to examples from similar categories than to examples from dissimilar categories. We develop a regularized kernel machine algorithm to use this category dependent similarity regularization. Our experiments on hundreds of categories show that our method can make significant improvement, especially for categories with no examples.
Cite
Text
Wang et al. "Comparative Object Similarity for Improved Recognition with Few or No Examples." IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2010. doi:10.1109/CVPR.2010.5539955Markdown
[Wang et al. "Comparative Object Similarity for Improved Recognition with Few or No Examples." IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2010.](https://mlanthology.org/cvpr/2010/wang2010cvpr-comparative/) doi:10.1109/CVPR.2010.5539955BibTeX
@inproceedings{wang2010cvpr-comparative,
title = {{Comparative Object Similarity for Improved Recognition with Few or No Examples}},
author = {Wang, Gang and Forsyth, David A. and Hoiem, Derek},
booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition},
year = {2010},
pages = {3525-3532},
doi = {10.1109/CVPR.2010.5539955},
url = {https://mlanthology.org/cvpr/2010/wang2010cvpr-comparative/}
}