Improving Weakly-Supervised Object Localization By Micro-Annotation
Alexander Kolesnikov and Christoph Lampert
Abstract
Weakly-supervised object localization methods tend to fail for object classes that consistently co-occur with the same background elements, e.g. trains on tracks. We propose a method to overcome these failures by adding a very small amount of model-specific additional annotation. The main idea is to cluster a deep networkÍs mid-level representations and assign object or distractor labels to each cluster. Experiments show substantially improved localization results on the challenging ILSVRC 2014 dataset for bounding box detection and the PASCAL VOC 2012 dataset for semantic segmentation.
Session
Posters 2
Files
Extended Abstract (PDF, 1M)
Paper (PDF, 9M)
DOI
10.5244/C.30.92
https://dx.doi.org/10.5244/C.30.92
Citation
Alexander Kolesnikov and Christoph Lampert. Improving Weakly-Supervised Object Localization By Micro-Annotation. In Richard C. Wilson, Edwin R. Hancock and William A. P. Smith, editors, Proceedings of the British Machine Vision Conference (BMVC), pages 92.1-92.12. BMVA Press, September 2016.
Bibtex
@inproceedings{BMVC2016_92,
title={Improving Weakly-Supervised Object Localization By Micro-Annotation},
author={Alexander Kolesnikov and Christoph Lampert},
year={2016},
month={September},
pages={92.1-92.12},
articleno={92},
numpages={12},
booktitle={Proceedings of the British Machine Vision Conference (BMVC)},
publisher={BMVA Press},
editor={Richard C. Wilson, Edwin R. Hancock and William A. P. Smith},
doi={10.5244/C.30.92},
isbn={1-901725-59-6},
url={https://dx.doi.org/10.5244/C.30.92}
}