Human and animal learning are naturally in the so-called semi-supervised mode of learning, i.e. with a limited teaching but more of self-training experiments. Human view is also often reinforced by multiple facets of learning experiences. Multi-view learning is concerned with the problem of machine learning from data represented by multiple distinct feature sets. The recent emergence of this learning mechanism is largely motivated by the property of data from real applications where examples are described by different feature sets or different “views”. In computer vision, low level visual features of images include colour, texture, shape, spatial relationships. There exist abundant and diverse researches in the content-based images/signal processing for information retrieval. For example, it is believed that Gabor filter resemble the performance of the mammalian visual cortical cells, in a sense of extracting features at different orientations and scales. Another successful example for texture classification of visual descriptor is local binary patterns. The aim of this project is to investigate efficient and accurate way that is scalable to large data sets without much labelled examples for extracting useful information. The project will be focus on the context of multi-view learning by combining known content-based visual features for image clustering/classification. This project will be focus on designing, developing algorithms, then validate and applying to computer vision applications such as visual surveillance, which demands efficient and powerful unsupervised and semi-supervised algorithms. Other example of applications is remote sensing images information retrieval.
UK honour equivalent in Computer Science, Maths, Engineering.