| Proj No. | A3201-251 |
| Title | Towards Agentic AI with Bio-inspired Camera-based Spatial Reasoning via Multi-modal Foundation Models |
| Summary | This research investigates a novel approach towards creating more agentic AI systems by focusing on bio-inspired camera-based spatial reasoning. Traditional computer vision relies on frame-based cameras, which can be inefficient and limited in dynamic environments. This project leverages bio-inspired cameras, such as event-based cameras, which provide asynchronous, high-temporal-resolution data streams mimicking aspects of biological visual systems. We hypothesize that by combining this event-based visual data with other sensory modalities (e.g., audio, inertial data) and processing it through powerful multi-modal foundation models, we can significantly enhance an AI agent's spatial reasoning capabilities. The project will explore different model architectures and fusion techniques for integrating these heterogeneous data streams, evaluating their performance on tasks requiring robust spatial understanding, such as navigation, object manipulation, and scene reconstruction. |
| Supervisor | Ast/P Wang Lin (Loc:S2 > S2 B2C > S2 B2C 91, Ext: +65 67905629) |
| Co-Supervisor | - |
| RI Co-Supervisor | - |
| Lab | Internet of Things Laboratory (Loc: S1-B4c-14, ext: 5470/5475) |
| Single/Group: | Single |
| Area: | Intelligent Systems and Control Engineering |
| ISP/RI/SMP/SCP?: |