Research programme

Research

Our work spans 5 connected areas, from multimedia safety and healthcare to the environment, visual scene understanding and trustworthy AI, with 32 publications so far.

05research areas

01

Multimedia Understanding and Safety

Fusing and reasoning across vision, audio and text to detect and localise hateful content, deepfakes and misinformation in online media, and building the benchmark datasets that measure this.

02

AI for Healthcare

Surgical video segmentation, medical image classification and radiology report generation, together with learning from brain and physiological signals for clinical and health applications.

03

AI for Environment

Monitoring oil spills, marine pollution and land-surface change in satellite and drone imagery, and deep learning for fast, transferable prediction of urban flooding.

04

Video and Scene Understanding

Anomaly detection in surveillance video and 3D point clouds, analysis of human pose, behaviour and facial reactions, and the robustness of visual perception models.

05

Trustworthy AI

Fairness across groups and tasks, learning from imbalanced and noisy data, and reliable, guaranteed evaluation with large language models.