CVPR '05 Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Volume 2 - Volume 02
CVPR '06 Proceedings of the 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition - Volume 1
Semi-supervised On-Line Boosting for Robust Tracking
ECCV '08 Proceedings of the 10th European Conference on Computer Vision: Part I
Robust Object Tracking with Online Multiple Instance Learning
IEEE Transactions on Pattern Analysis and Machine Intelligence
Multiple Object Tracking Using K-Shortest Paths Optimization
IEEE Transactions on Pattern Analysis and Machine Intelligence
FlowBoost -- Appearance learning from sparsely annotated video
CVPR '11 Proceedings of the 2011 IEEE Conference on Computer Vision and Pattern Recognition
| Hi-index | 0.00 |
A recent trend in object detection and tracking is using multiple-instance learning (MIL) to resolve the uncertainties in the training set. Though using online multiple instance learning instead of traditional instance-based learning can lead to a more robust appearance classier, but it also tends to drift or fail in case of wrong updates during the online self-learning process. In this work we propose a method to combine the benefit of online MIL learning and off-line/batch learning to get a robust appearance model which is able to effectively handle drifting problem. Our method not only copes with ambiguity with power of multiple instance learning, but also uses off-line learning with a sample weights descending in a iterative framework to suppress drifting in the result of online MIL. We demonstrate the effectiveness and robustness of our method on several challenging video clips and show performance improvement comparing to other state-of-art approaches especially to online MIL learning in a fully occlusion scene.