The development of deep learning CNN architectures has led to remarkable progress in the field of computer vision. As a consequence, successful systems are now available for real world applications in a variety of areas including visual surveillance, sports analysis, gaming and security. However, variability in human shapes and faces, appearances, postures and personalities makes the automatic recognition of a specific individual a very challenging task especially when they are either non-cooperative, i.e. they are not aware of the presence of a computer vision system, or uncooperative, i.e. trying to prevent a system to recognise them. This is exemplified by the embarrassing failure of the face recognition system recently tested by the Metropolitan police at the Notting Hill carnival in 2017 which was largely reported in the media (https://www.scientificamerican.com/article/dna-techniques-could-transform-facial-recognition-technology/
This project proposes to address person recognition from either images or videos by developing a novel two-stream CNN-Profile HMM architecture, where the successful deep learning CNN architecture which is designed to optimise inter individual discrimination is enhanced by intra individual variation models. Inspired by extraordinary individuals, the so-called super recognisers, who have the ability of accurately identifying individuals thanks to both an excellent photographic memory and capability to extrapolate from several viewpoints, it is proposed to create a canonical model for each individual of interest by taking advantage of existing imagery representing them. Such models will be created using the novel ‘vide-omics’ paradigm which has recently been developed at Kingston University . Founded on the principles of genomics where variability is the expected norm rather than an inconvenience to control, ‘vide-omics’ will allow modelling variations of a human’s personal characteristics, such as face, appearance and gait, by interpreting them as the product of mutations applied to a canonical model, the common ancestor of all representations of that individual. Similarly to genomics where models of gene families are produced by encoding mutations between each member using Profile Hidden Markov Models (P-HMMs), novel individual descriptors will be developed based on P-HMMs to model quantitatively known variations. The suitability of P-HMMs to encode object variations in images has already been demonstrated in a recent publication .
Applicants should have, at least, an Honours Degree at 2.1 or above (or equivalent) in Computer Science or related disciplines. In addition, they should have excellent programming skills in Matlab, Java and/or C++ and an interest in machine learning.
Qualified applicants are strongly encouraged to informally contact the supervising academic, Dr Nebel ([email protected]
), to discuss the application. More on Dr Nebel’s research group and activities can be found on his personal website: https://sites.google.com/site/jeanchristophenebel/
 Vide-omics: A Genomics-inspired Paradigm for Video Analysis, I. Kazantzidis, F. Florez-Revuelta, M. Dequidt, N. Hill and J.-C. Nebel, Computer Vision and Image Understanding, 166:28-40, 2018
 Profile Hidden Markov Models for Foreground Modelling, I. Kazantzidis, F. Florez-Revuelta and J.-C. Nebel, IEEE International Conference on Image Processing (ICIP 2018), Athens, Greece, October 7-10, 2018