Department of Computer Engineering
S E M I N A R
Utilizing Multiple Instance Learning for Computer Vision Tasks
Computer Engineering Department
The Multiple Instance Learning (MIL) problem arises in many application domains, whereas it is particularly suitable for computer vision problems due to the difficulty of obtaining manual labeling. Multiple Instance Learning methods have large applicability to a variety of challenging learning problems in computer vision, including object recognition and detection, tracking, image classification, scene classification and more.
Multiple Instance Learning operates over bags of instances, as opposed to single instances in standard supervised learning. A bag is positively labeled if it contains at least one positive instance; otherwise it is labeled as a negative bag. The task is to learn some concept from the training set for unseen bags. A vital component of using Multiple Instance Learning in computer vision is its design for abstracting the visual problem to multi-instance representation that is, determining what the bag is and what are the instances in the bag.
In this context, we consider three different computer vision problems and propose solutions with novel representations. Firstly for image retrieval; We look into this problem of image re-ranking and propose a method that automatically constructs multiple candidate bags, which are likely to contain relevant images. Secondly for recognizing actions from still images; we extract several candidate object regions and approach the problem of identifying related objects from a weakly supervised point of view. Finally for interaction recognition from videos; videos are composed of irrelevant frames roughly starts from a standing pose before interaction and ends with a standing pose after interaction. To overcome this problem, we use the idea of Multiple Instance Learning to tackle irrelevant actions in whole sequence classification.
Experimental results on data sets verify the performance of the proposed algorithms are provided.
DATE: 4 March, 2013, Monday @ 15:40