About the Project
Deep Learning is a fast-growing area in the field of machine learning within which systems can be trained to identify patterns in a way that loosely represent biological neural computation. Recent advances in such Deep Neural Network technology have led to a significant performance leap in pattern-recognition tasks such as computer vision (image understanding), text mining (language understanding) and voice recognition (audio understanding). This advancement has been further enhanced by the evolution of low-cost Graphical Processing Units into massively-parallel computing platforms capable of processing vast datasets for such tasks.
More recent advances, including leading work at Durham on the GANomaly algorithms, has seen a rise in the use of semi-supervised and unsupervised techniques. These approaches address problems where constructing large, annotated training datasets maybe expensive, time-consuming or where all scenarios cannot be readily covered prior to deployment. A particular area of research interest is automated anomaly detection where the set of potential anomalies may include rare or even unforeseen occurrences for which example images cannot be obtained. Within this area, techniques such as GANomaly allow anomaly detection to be performed using only non-anomolous (normal) image samples which are often in abundance compared to the scarcity of anomaly samples.
Building on world-leading work at Durham on automated anomaly detection, this project focuses on the advancement of these techniques to anomaly detection within both 3D and 4D (3D spatial + time) imagery. Advancing the prior work on this topic from conventional (2D still/video) imagery, will see anomaly detection applied to cutting edge ultrasound, CT and MRI scan data sets for applications within both medicine (COSMONiO) and aviation/border security (Durham).
COSMONiO, an SME with bases in the UK and the Netherlands, designs cutting-edge computer-vision and machine-learning systems that automate the process of extracting visual information from images and other data sources even under the most challenging conditions (www.cosmonio.com).
The successful applicant will be based in the Department of Computer Science of Durham University – ranked in the top 5 for both Engineering & Computer Science (Complete University Guide 2020, http://www.durham.ac.uk/computer.science/).
The project will take the form of an industrially supported studentship with the student primarily based at Durham University whilst working for short industrial placement periods with Cosmonio throughout the PhD.
An overview of related research work within the Durham team is available from http://www.durham.ac.uk/toby.breckon/publications/ and https://cwkx.github.io/ which illustrates both work across the automotive sensing, medical imaging, surveillance and security screening projects within the group.
Applicants should have good background in any of computer science, artificial intelligence, engineering, physics or a related discipline with a strong programming ability in a high level language (preferably C/C++, Java or Python) and a highly competent mathematical background is essential (especially algebra, statistics and geometry). Candidates should hold at least a 2:1 honours degree or equivalent (Masters degree a plus). Prior experience in computer vision and machine learning is a plus although not essential. EU students would also need to provide evidence of English language competency.
This studentship is supported through the Intensive Industrial Innovation Programme, part funded by the European Regional Development Fund. The studentship covers full Home fees and a tax-free stipend at the standard EPSRC rate with an additional bursary top-up from the industrial sponsor (£18,285, 2020/21).
Applicants from outside the UK/EU are not eligible for this award.
1 October 2020
Informal enquiries can be made to either Prof. Toby Breckon ([email protected]) or Dr Chris Willcocks ([email protected]).
How to apply:
Please visit the University applications page at https://www.dur.ac.uk/study/pg/apply/ (specify project title: “Real-time Anomaly Detection in 3D and 4D Data using Deep Machine Learning”, supervisor: Prof. Toby Breckon / Dr. Chris Willcocks, Computer Science)
31 August 2020. (for guidance as decisions will be made on applicants when received).
Why not add a message here
Based on your current searches we recommend the following search filters.
Based on your current search criteria we thought you might be interested in these.