MrNeRF (@janusch_patas)
2024-10-11 | โค๏ธ 367 | ๐ 68
DifFRelight: Diffusion-Based Facial Performance Relighting
Fun fact: First 3DGS paper from Netflix!
Contributions (cited): (1) A novel framework for facial performance relighting that trains on a subject-specific paired dataset and generalizes to novel lighting conditions, enabling relighting from free viewpoints and unseen facial expressions
(2) A diffusion-based relighting model that spatially conditions the flat-lit input image, utilizing lighting information as global control to generate high-quality relit results.
(3) A scalable dynamic 3DGS technique for reconstructing long sequences, ensuring temporal consistency in flat-lit inputs for coherent inference by the relighting model.
(4) A unified lighting control that combines a new area lighting representation with directional lighting, offering versatile lighting controls as well as enabling the composition of complex environment lighting
๋ฏธ๋์ด
