Multiple-instance image ranking for sketch-based image retrieval

Abstract

There are image classification problems in which each image is represented by regions of interest; for each region a series of features can be extracted. As a result a set of feature vectors are available, and it is necessary to assign a label to this set of instances. The Multiple-Instance Learning (MIL) studies the problem in which each object is described as a bag (set of instances). This project aims to study the properties of MIL and develop solutions to image ranking for image retrieval using this approach; more specifically, we want to investigate this approach on sketch-based image retrieval (SBIR) tasks. We seek a solution that suits each query accordingly by performing instance selection on the bag's instances (that represent each image on the training sample) and presenting those based on a relevance ranking model based on the current query. Therefore, the following important aspects inherent to this project are: Multiple Instance Learning, feature extraction over sketches, image ranking and the instance selection. (AU)