Associate Professor, Hong Kong Baptist University
2 papers at NeurIPS 2025
A semi-supervised learning approach for high-fidelity spectral reconstruction to deal with the metameric dilemma
The proposed Compositional Phase Diffusion framework consistently generates semantically aligned multi-clip motion with smooth transitions by using latent-phase diffusion modules (SPDM and TPDM) to preserve phase continuity and enable inbetweening.
