Human Actions Analysis - Videos
Overview
This project is a collaborative effort with Fennex Ltd to develop AI-driven solutions for proactive safety enhancement through video understanding. The core objective is to predict potential hazards by recognising and interpreting human activities in video data, enabling early anticipation of unsafe situations and supporting timely interventions in real-world industrial setting.
Research Areas
Building on this foundation, the project has been extended to the domain of surgical video analysis. We are developing advanced methods to understand fine-grained surgical actions and to predict the surgeon’s next possible actions during a procedure. By anticipating critical steps and potential deviations, the approach aims to improve situational awareness, reduce risk, and enhance overall safety in the operating room.
Collaborators
- Fennex Limited
- School of Computing, Engineering and Technology, Robert Gordon University (SoCET)
Funding
The work was funded by Innovate UK and Fennex Limited £207,835.00. And a fully-funded PhD studentship by SoCET.
Selected Publications
Shuvo, M.R., Mekala, M.S. & Elyan, E. Deep Learning and Attention-Based Methods for Human Activity Recognition and Anticipation: A Comprehensive Review. Cogn Comput 17, 158 (2025). https://doi.org/10.1007/s12559-025-10513-2
Shuvo, M.R., Mekala, M.S. & Elyan, E, MSBATN: Multi-Stage Boundary-Aware Transformer Network for action segmentation in untrimmed surgical videos, Computer Vision and Image Understanding, Volume 261, 2025, 104521, https://doi.org/10.1016/j.cviu.2025.104521.
Team
The team includes two academics, one PhD student, and one Postdoc researcher.