In defence of negative mining for annotating weakly labelled data

Authors:
Parthipan Siva;Chris Russell;Tao Xiang
Affiliations:
Queen Mary, University of London, London, United Kingdom;Queen Mary, University of London, London, United Kingdom;Queen Mary, University of London, London, United Kingdom
Venue:
ECCV'12 Proceedings of the 12th European conference on Computer Vision - Volume Part III
Year:
2012

Citing 11
Cited 1

A framework for multiple-instance learning

NIPS '97 Proceedings of the 1997 conference on Advances in neural information processing systems 10
Space-time Interest Points

ICCV '03 Proceedings of the Ninth IEEE International Conference on Computer Vision - Volume 2
MILES: Multiple-Instance Learning via Embedded Instance Selection

IEEE Transactions on Pattern Analysis and Machine Intelligence
The Pascal Visual Object Classes (VOC) Challenge

International Journal of Computer Vision
A survey on vision-based human action recognition

Image and Vision Computing
Object Detection with Discriminatively Trained Part-Based Models

IEEE Transactions on Pattern Analysis and Machine Intelligence
Localizing objects while learning their appearance

ECCV'10 Proceedings of the 11th European conference on Computer vision: Part IV
A survey of vision-based methods for action representation, segmentation and recognition

Computer Vision and Image Understanding
MILIS: Multiple Instance Learning with Instance Selection

IEEE Transactions on Pattern Analysis and Machine Intelligence
Weakly supervised object detector learning with model drift detection

ICCV '11 Proceedings of the 2011 International Conference on Computer Vision
Scene recognition and weakly supervised object localization with deformable part-based models

ICCV '11 Proceedings of the 2011 International Conference on Computer Vision

Learning discriminative localization from weakly labeled data

Pattern Recognition

Quantified Score

Hi-index	0.00

Visualization

Abstract

We propose a novel approach to annotating weakly labelled data. In contrast to many existing approaches that perform annotation by seeking clusters of self-similar exemplars (minimising intra-class variance), we perform image annotation by selecting exemplars that have never occurred before in the much larger, and strongly annotated, negative training set (maximising inter-class variance). Compared to existing methods, our approach is fast, robust, and obtains state of the art results on two challenging data-sets --- voc2007 (all poses), and the msr2 action data-set, where we obtain a 10% increase. Moreover, this use of negative mining complements existing methods, that seek to minimize the intra-class variance, and can be readily integrated with many of them.