Post

SwiMDiff Scene-wide Matching Contrastive Learning with Diffusion Constraint for Remote Sensing Image

  • 论文名称: SwiMDiff: Scene-wide Matching Contrastive Learning with Diffusion Constraint for Remote Sensing Image
  • Link: https://arxiv.org/abs/2401.05093
  • Published in: Arxiv 2024
  • Type: Pretrain
  • Code/Project: —
  • 备注: 对比学习,用同一场景下的图片构建了一个FNS,并用一个基于Diffussion的图像恢复任务做辅助任务
  • Backbone: Resnet
  • Short Summary: 以往的CL忽略了细粒度特征和负样本中的潜在联系,SwiMDiff采用场景范围的匹配方法,有效地重新校准标签,将来自同一场景的数据识别为假负样本。此调整使CL更适用于遥感的细微差别。此外,SwiMDiff将CL与扩散模型无缝集成。通过实施像素级别的扩散约束,我们增强了编码器捕捉图像的全局语义信息和细粒度特征的能力。
This post is licensed under CC BY 4.0 by the author.