DINOv2: Learning Robust Visual Features without Supervision

  • 类型:arxiv
  • 标识:2304.07193
  • 链接:https://arxiv.org/abs/2304.07193
  • 主题:rag
  • 主分类:multimodal
  • 形态:method
  • 被引:9976
  • 被引来源:Semantic Scholar
  • S2被引:9976
  • OpenAlex被引:1067
  • 影响力被引:1445
  • TLDR:This work revisits existing approaches and combines different techniques to scale the pretraining in terms of data and model size, and proposes an automatic pipeline to build a dedicated, diverse, and curated image dataset instead of uncurated data, as typically done in the self-supervised literature.
  • OpenAlex ID:W4366208220
  • OpenAlex DOI:10.48550/arxiv.2304.07193
  • DOI:10.48550/arxiv.2304.07193
  • DOI来源:OpenAlex
  • 开放获取:green
  • 开放获取链接:https://arxiv.org/pdf/2304.07193
  • OpenAlex更新:2026-08-22
  • 待LLM分类:否
  • 标题中文:DINOv2:无监督学习鲁棒的视觉特征
  • TLDR中文:本文回顾现有方法,并融合多种技术从数据与模型规模两方面扩展预训练,提出一条自动化流水线以构建专用、多样且经过筛选的图像数据集,替代自监督文献中常用的未筛选数据。
  • 来源文件
  • [OpenAlex discover]
  • [S2 enrich]