This work reformulates relighting as an illumination feature transport problem and introduces Consistent Feature Transport (CFT), a training principle that explicitly enforces illumination-consistent transport between source and target image distributions.
Bohan Zhang, Huan Liang, Yuhang He et al.· arXiv.org· 0 citations
StructGen is proposed, which employs a structured, dictionary-like format to encode multiple reference images, thereby enabling explicit and unambiguous specification of generation intentions, and consistently outperforms existing methods on both semantic alignment and detailed reference-generation consistency.
MIDAS is presented, an interview-aware multimodal framework that preserves the question-response structure of clinical interviews for depression assessment and integrates linguistic, acoustic, and visual-behavioral representations through gated multimodal fusion and shared large language model-based contextual modeling...