Throughout 2025, AI-PRISM partners consolidated their position at the forefront of computer vision, robotics and human-centred AI research. This year’s contributions demonstrate a clear evolution: advancing state-of-the-art perception methods while strengthening industrial validation, datasets and responsible AI integration.
AI-PRISM research featured prominently at the IEEE/CVF Winter Conference on Applications of Computer Vision 2025, the British Machine Vision Conference 2025, the IEEE/CVF Conference on Computer Vision and Pattern Recognition 2025, and the Applied Human Factors and Ergonomics Conference 2025, as well as MDPI Telecom events. The presented work addressed high-resolution open-vocabulary 6D pose estimation, efficient zero-shot pose estimation using frozen foundation models, 3D part segmentation through geometric aggregation of visual features, and functionality understanding in complex 3D scenes. The introduction of the CHIP multi-sensor dataset and a dataset capturing worker operation durations further reinforced the project’s commitment to reproducibility and real industrial relevance.
Peer-reviewed journal publications in Open Research Europe, IEEE Access, and IEEE Transactions on Pattern Analysis and Machine Intelligence expanded the scope beyond technical perception systems. Research explored large language models in human-robot collaboration, cognitive validation against context-induced hallucinations, ethical AI-enabled HRC in manufacturing, and the performance of Wi-Fi standards in industrial environments.
Taken together, the 2025 scientific outputs reflect AI-PRISM’s integrated approach: combining cutting-edge AI perception, robust datasets and foundation models with ethical, cognitive and industrial considerations to advance sustainable and trustworthy human-robot collaboration.
Read our research results here: