Output details
11 - Computer Science and Informatics
University of Westminster
Region analysis through close contour transformation using growing neural gas
<23>Originality: This paper introduces a new discriminative shape descriptor that generalises to any contour and region, and can be used to speed up the modelling and the analysis of any shape.
Significance: The paper proposes a solution to a fundamental problem of extracting semantically meaningful visual regions. It has also established new partnerships with Orange (France), and the consulting company ClopiNet (USA), in the organisation of workshops on Active and Autonomous Learning.
Rigour: Mathematical and Algorithmic analyses support all major findings of this research. Results published in a leading, peer-reviewed conference.