Computer Vision & Autonomous Systems
The Computer Vision and Autonomous Systems (CVAS) group advances AI-driven perception, robotics and computational imaging to develop autonomous systems that understand dynamic, multimodal environments, act within them and collaborate with humans. Research spans scene understanding, robotic perception and manipulation, generative and probabilistic AI, and human motion analysis, with applications in engineering, healthcare and manufacturing.
Our scene understanding research addresses urban, industrial and natural environments, including autonomous driving, biodiversity monitoring (NERC-funded DEAL) and anomaly detection in additive manufacturing. We develop deep learning approaches to 3D vision, contextual reasoning and decentralised learning, including under limited or noisy supervision.
Robotics research focuses on systems capable of adapting to dynamic environments without exhaustive pre-programming. Projects include deformable object manipulation, Scotland–Japan teleoperation and AI-driven assembly with HAL Robotics and the National Manufacturing Institute Scotland.
In healthcare, computer vision and probabilistic machine learning support clinical decisions through projects on infant neurological assessment (SAFEVID-TREs; BrAInwaves), epileptic spasm detection (VISDOM-AI), brain tumour treatment response (BIOGRAPH) and bone segmentation for surgical planning. Our generative AI and Bayesian research also supports scientific discovery, including virtual unrolling of ancient Herculaneum scrolls, recognised by the 2025 Vesuvius Challenge Auto Segmentation Prize.
Project Highlights
Computer Vision
- DEAL: Decentralised Learning for automated image analysis and biodiversity monitoring (NERC, 2024–2027)
- Virtual unwrapping of the Herculaneum/Vesuvius Scrolls (Vesuvius Challenge Auto Segmentation Prize, 2025)
- High-fidelity capture of 3D surfaces using neural representations (Royal Society Research Grant, 2022–2023)
Robotics & Autonomous Systems
- Anomaly detection in additive manufacturing, in partnership with HAL Robotics (Royal Commission for the Exhibition of 1851 Industrial Fellowship, 2024–2028)
- Robotic teleoperation: Examining non-expert user experiences in Scotland and Japan (Royal Society of Edinburgh, 2024)
Healthcare & Clinical AI
- SAFEVID-TREs: Spasms analysis using federated learning from videos across multiple Trusted Research Environments (MRC, 2026–2027)
- VISDOM-AI: Video-based infantile epileptic spasm detection from body movements (Epilepsy Research Institute Endeavour Grant, 2025–2027)
- BrAInwaves: Automating General Movement Assessment for early prediction of Cerebral Palsy (RS Macdonald Charitable Trust, 2026–2027)
- Infant Movement Assessment for neurological condition identification from videos using cloud-based AI (EPSRC/MRC IAA, 2024–2025)
- BIOGRAPH: Understanding and predicting brain tumour treatment response using AI (Brain Research UK, 2025–2028)
- Machine Learning Driven Bone Segmentation for Fracture Surgery Planning (Glasgow Innovation Cluster Fund, 2025)