Skip to main content
School of Mathematical Sciences

Data-driven Image Processing Methods with Applications to Wildlife Conservation

  • Supervisor: Dr Kostas Papafitsoros
  • Funding: EPSRC PhD Studentship
  • Deadline: 1st December 2023 for an April 2024 start

Project description:

Facilitated by the advances of imaging technologies, as well as the nowadays wide availability of data, wildlife conservation projects are increasingly using large scale imaging datasets to answer important ecological questions, towards preservation of endangered species. The majority of such projects crucially rely on Photo-Identification (Photo-Id), which is the process of identifying individual animals, by using their unique external morphological patterns (e.g. stripes in zebras, facial scales in sea turtles). However, this task is often hindered by the low quality and resolution of the images, which is due to bad lighting conditions, large distance between animals and cameras, elusive individual animals etc.

The aim of this PhD project is to develop, analyse and implement data-driven mathematical image processing methods, in order to enhance low quality images of individual animals, facilitating their identification. In particular, data-driven super-resolution techniques will be considered, with the objective being to produce high resolution images from low resolution ones, recovering crucial identifying information, which is not initially detectable. This constitutes a highly ill-posed inverse problem which asks for tailored regularisation (prior information imposed to the reconstructed image). The project aims to develop deep learning-based image priors (neural network-based regularisation functionals), that are trained with the help of large databases of the focal species. It will investigate methods of incorporating this type of regularisation into the reconstruction process, e.g. through generative modelling, plug-and-play priors, adversarial regularisers and others. The project will be supported by an existing comprehensive photo-database of loggerhead sea turtles, a benchmark species for animal Photo-Id. This is a unique opportunity to apply modern data-driven imaging methods, to benefit the study and conservation of endangered species.

The ideal candidate should possess a masters degree, or equivalent, in applied mathematics and also familiar with deep learning, imaging and computer vision techniques. Strong programming skills (e.g. Python) are essential. 

Funding

The studentship is funded by EPSRC PhD Studentship and will cover tuition fees, and provide an annual tax-free maintenance allowance for 3.5 years at the UKRI rate (£20,622 in 2023/24).

These studentships are open to those with Home and International fee status; however, the number of students with International fee status which can be recruited is capped according to the EPSRC terms and conditions so competition for International places is particularly strong.  Tuition fee rates can be found at: https://www.qmul.ac.uk/postgraduate/research/funding_phd/tuition-fees/ 

Further information:

How to apply

Entry requirements

Fees and funding

Back to top