Purify-then-Align: Towards Robust Human Sensing under Modality Missing with Knowledge Distillation from Noisy Multimodal Teacher
在模态缺失情况下通过知识蒸馏从噪声多模态教师中实现鲁棒的人体感知:Purify-then-Align
机构 * School of Software Engineering, Xi’an Jiaotong University(西安交通大学软件工程学院) ; Institute of Computer Science, Universität Bern(伯尔尼大学计算机科学研究所) ; College of Computing and Data Science, Nanyang Technological University(南洋理工大学计算与数据科学学院)
专题命中 多模态生成 :multimodal(title,abstract);cross-modal(abstract);分类 cs.CV;any-to-any(comments)
AI总结 本文提出PTA框架,通过元学习和知识扩散解决多模态人体感知中模态缺失问题,通过净化知识源和对齐模态提升单模态模型鲁棒性。
Comments Accepted by CVPR 2026 Workshop On Any-to-Any Multimodal Learning