PhD Studentship in Deep Unsupervised and Active Learning Approaches to Satellite Image Analysis University of Southampton School of Electronics and Computer Science
|
PhD Studentship in Deep Unsupervised and Active Learning Approaches to Satellite Image Analysis Agents, Interactions & Complexity Location: Highfield Campus Closing Date: Monday 31 August 2020 Reference: 1281520FP Supervisory Team: Jonathon Hare, Adam Prügel-Bennett and Patrick Osborne Project description The Vision, Learning & Control group in the School of Electronics and Computer Science, and School of Geography and Environmental Science, at the University of Southampton, together with Ordnance Survey, invite applications for a PhD to investigate recent advances in deep learning and their application to satellite image interpretation. This project builds on a strong relationship between all three groups and will combine novel research with real-world environmental applications. Satellite data are essential for many regional, national and international monitoring and mapping problems. Excellent analytical results have been achieved in a diverse range of applications such as monitoring environmental conditions to mapping urban change. Most of these applications rely on careful adaptation and tuning of a range of processing techniques to the specific geographic region or problem domain. This ambitious and timely PhD research project aims to minimise the amount of over-specialisation of the image processing and instead build a more generic approach by developing a method for pre-processing massive (up to global) data sets in a way that is generic to the final application. This research will build on recent advances in unsupervised and active learning approaches and apply them to pre-training with satellite data. We postulate that, given a large enough satellite imagery data set, methods to draw out the most informative components of the data can be learned, in particular using deep learning techniques. These components, or representations of the imagery, can then be applied to a range of real-world problems. The successful applicant is likely to have a good (at least 2.1) first degree in Computer Science, Mathematics, Physics or in Geography/Environmental Science with strong quantitative components in GIS and remote sensing. A sound background in programming, ideally in Python, is essential and experience of GPU accelerated deep learning frameworks is desirable. Equality, diversity and Inclusion is central to the ethos in ECS. Applications are especially welcomed from underrepresented communities. The candidate will be based within at the University of Southampton's Highfield campus but will also spend time at Ordnance Survey's offices in Southampton. If you wish to discuss any details of the project informally, please contact Dr Jonathon Hare, VLC Research Group, Email: jsh2@ecs.soton.ac.uk, Tel: +44 (0) 2380 59 7678. Entry Requirements A very good undergraduate degree (at least a UK 2:1 honours degree, or its international equivalent). Closing date: applications should be received no later than 31 August 2020 for standard admissions, but later applications may be considered depending on the funds remaining in place. Funding: full tuition fees for EU/UK students plus for UK students, an enhanced stipend of £15,285 tax-free per annum for up to 3.5 years. How To Apply Applications should be made online, please select the academic session 2020-21 “PhD Computer Science (Full time)” as the programme. Please enter Jon Hare under the proposed supervisor. Applications should include: Research Proposal Curriculum Vitae Two reference letters Degree Transcripts to date Apply online: For further information please contact: feps-pgr-apply@soton.ac.uk
|
