Yangyang at ICCV 2025

On 21 October 2025, Yangyang Xu's work was presented at the IEEE/CVF International Conference on Computer Vision (ICCV 2025), held in Honolulu, Hawaii, from 19 to 23 October 2025. ICCV is one of the premier global conferences in computer vision, bringing together leading researchers and practitioners to discuss the latest advances in visual intelligence.

Yangyang’s paper, titled “Cross-Subject Mind Decoding from Inaccurate Representations”, introduces the Bidirectional Autoencoder Intertwining (BAI) framework. This novel approach leverages fMRI signals from individuals to reconstruct real-world images they have visualised, effectively addressing traditional challenges such as inter-subject variability and error accumulation. By unifying subject data and refining representations, the BAI framework achieves state-of-the-art performance, high-fidelity reconstructions, and adapts to new subjects with minimal data. For more details, please refer to the paper.