AnyMod-LLVE: Low-Light Video Enhancement with Modality-Agnostic Inference
AnyMod-LLVE: 模态无关推理的低光照视频增强
机构 * University of Science and Technology of China(中国科学技术大学)
AI总结 提出AMNet统一多模态框架,通过空间-频谱双门控转换器学习辅助模态与RGB输入的对应关系,支持推理时任意模态组合,解决低光照视频增强中辅助模态缺失问题。
Comments Accepted at ICML 2026; Project page and code: https://lhfgghc.github.io/LLVE-AMNet