arXivDaily arXiv每日学术速递 周一至周五更新

期刊&会议

International Journal of Computer Vision · 期刊 · Computer Vision

2026-01-13 至 2026-01-13 共收录 2
2601.07117 2026-01-13 cs.CV cs.AI

Few-shot Class-Incremental Learning via Generative Co-Memory Regularization

基于生成共记忆正则化的少样本类增量学习

Kexin Bao, Yong Li, Dan Zeng, Shiming Ge

机构 * Institute of Information Engineering(信息工程研究所) Chinese Academy of Sciences(中国科学院) School of Cyber Security(网络安全学院) University of Chinese Academy of Sciences(中国科学院大学) Department of Communication Engineering(通信工程系)

AI总结 本文提出基于生成共记忆正则化的少样本类增量学习方法,通过微调生成编码器和构建类记忆来提升模型在少量样本下的识别准确率,同时减少灾难性遗忘和过拟合。

Comments Accepted by International Journal on Computer Vision (IJCV)

详情

展开后加载摘要…

URL PDF HTML 收藏
2506.04115 2026-01-13 cs.CV

Multi-view Surface Reconstruction Using Normal and Reflectance Cues

多视角表面重建使用法线和反照率线索

Robin Bruneau, Baptiste Brument, Yvain Quéau, Jean Mélou, François Bernard Lauze, Jean-Denis Durou, Lilian Calvet

机构 * University of Zurich(苏黎世大学) Université de Toulouse(图卢兹大学) CNRS, UNICAEN, ENSICAEN, Normandie Université(法国国家科学研究中心、UNICAEN、ENSICAEN、诺曼底大学) FittingBox(FittingBox公司) University of Copenhagen(哥本哈根大学)

AI总结 本文提出了一种结合多视角法线和反照率地图的表面重建框架,通过参数化辐射向量实现高保真3D重建,尤其在复杂材质和可见性挑战下表现优异。

Comments 22 pages, 15 figures, 11 tables. Accepted to IJCV. A thorough qualitative and quantitive study is available in the supplementary material at https://drive.google.com/file/d/1KDfCKediXNP5Os954TL_QldaUWS0nKcD/view?usp=drive_link. The project page can be accessed via https://robinbruneau.github.io/publications/rnb_neus2.html. The source code is available at https://github.com/RobinBruneau/RNb-NeuS2

详情

展开后加载摘要…

URL PDF HTML 收藏