Primary Supervisor –Professor Michal Mackiewicz

Scientific background

Marine litter poses a major threat to ocean health and livelihoods. Hence, new scalable, automated methods for data collection and analysis are needed to enhance our understanding of litter sources, pathways and impact. Cefas is developing a deep learning (DL) algorithm for visible light (VL) and has collected a large training data set for the 89 litter classes. However, there is recognition of the need for multispectral images to enhance the accuracy of the algorithms being developed when distinguishing material type. Thus, Cephas is developing a new laboratory to help characterize the multispectral reflectance of materials.

Research methodology

The student will use the existing VL database for key materials, but more importantly will also collect multispectral data through an improved laboratory setup with the aim of training DL algorithms. Importantly, the algorithms developed will need to be robust to real-world lighting changes and for long-term use, likely with imaging hardware that did not exist during development. This will require an approach that takes into account the physics of multispectral image formation, including three key variables: the spectral sensitivities of the sensor, the variable daylight spectrum, and a wide range of reflectance spectra of the relevant materials.

Goals

Development of a multispectral imaging dataset for marine litter materials by expanding the existing VL dataset.

Design and evaluate DL models capable of classifying marine litter types using multispectral data, with an emphasis on achieving robustness with different spectral channel configurations and lighting conditions.

Implement and validate device-independent representations. Investigate domain adaptation and transfer learning techniques to develop models that generalize across different imaging devices, including future sensors with unknown spectral sensitivities.

an exercise

The student will be based in the Color and Imaging Lab in the College of Computer Science with experience in designing and evaluating imaging solutions and will have the opportunity to work with scientists and engineers at Cefas. They will conduct training specific to this project including principles of photography, laboratory measurement, computer vision, ArcGIS, potential field work, and drone flight training.

Person specifications

Experience and/or enthusiastic interest in one or more of the following areas: environmental monitoring, artificial intelligence, computer vision, or multispectral imaging.

Entry requirements

At least a UK BA (Hons) equivalency of 2:1. English language requirements (Faculty of Science equivalent: IELTS 6.5 overall, 6 in each category).

Acceptable first degree: Computer Science/Physics/Mathematics or other computational disciplines.

Study method

Full time

Start date

October 1, 2026

Financing information

ARIES Scholarships are governed by UKRITerms and Conditions. Successful candidates who meet UKRI’s eligibility criteria will be awarded a fully-funded scholarship, which covers fees, maintenance stipends (20,780 per annum for 2025/26) and a Research Training and Support Grant (RTSG). A limited number of scholarships are available to international applicants, with the difference between ‘national’ and ‘international’ fees waived by the registered university. However, please note that ARIES funding does not cover additional costs associated with moving to and living in the UK, such as visa costs or additional health fees.