DiRe: Diversity-promoting Regularization for Dataset Condensation
DiRe:促进多样性的正则化方法用于数据集压缩
机构 * Department of Artificial Intelligence, Indian Institute of Technology Hyderabad(人工智能系,印度理工学院海得拉巴) ; Centre for Responsible AI, Wadhwani School of Data Science & AI, Indian Institute of Technology Madras(负责任人工智能中心,Wadhwani数据科学与人工智能学院,印度理工学院马德拉斯)
AI总结 DiRe通过引入余弦相似度和欧几里得距离的正则化方法,提升数据集压缩的多样性与泛化能力。
Comments Accepted at WACV 2026. v2: Optimized figure assets to reduce PDF size, no content changes