Jove
Visualize
联系我们
JoVE
x logofacebook logolinkedin logoyoutube logo
关于 JoVE
概览领导团队博客JoVE 帮助中心
作者
出版流程编辑委员会范围与政策同行评审常见问题投稿
图书馆员
用户评价订阅访问资源图书馆顾问委员会常见问题
研究
JoVE JournalMethods CollectionsJoVE Encyclopedia of Experiments存档
教育
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab Manual教师资源中心教师网站
使用条款与条件
隐私政策
政策

相关实验视频

Updated: Mar 15, 2026

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

1.2K

SSRT-DETR:域自适应性半监督探测器

Wenshuai Zhang1, Dong Zhou1, Wenjie Xie1

  • 1The Research Institute of Electronic Science and Technology, University of Electronic Science and Technology of China, Chengdu 610054, China.

Sensors (Basel, Switzerland)
|March 14, 2026
PubMed
概括

本研究介绍了SSRT-DETR,这是一个新的半监督,域适应性对象检测框架. 它通过改善对域转移的匹配和伪标签策略来提高具有挑战性的数据集的性能.

相关概念视频

您也可能阅读

相关文章

通过共同作者、期刊和引用图与本文相关的文章。

排序
Same author

Group percolation in interdependent networks.

Physical review. E·2018
Same author

Network recovery based on system crash early warning in a cascading failure model.

Scientific reports·2018
Same author

Late-onset anti-N-methyl-d-aspartate receptor encephalitis in China.

Epilepsy & behavior : E&B·2018
Same author

Wireless Phone Use and Risk of Adult Glioma: Evidence from a Meta-Analysis.

World neurosurgery·2018
Same author

Progressive cystic lesion in a middle-aged patient with tuberous sclerosis complex: A case report.

Medicine·2018
Same author

The focal alteration and causal connectivity in children with new-onset benign epilepsy with centrotemporal spikes.

Scientific reports·2018

科学领域:

  • 计算机视觉 计算机视觉
  • 机器学习 机器学习
  • 人工智能的人工智能

背景情况:

  • 域自适应对象检测对于现实应用至关重要,但在域移动时面临挑战.
  • 现有的方法与匈牙利匹配灵敏度和固定的伪标签值扎,特别是与阶级不平衡和场景变化.

研究的目的:

  • 开发一个强大的半监督,域适应物体检测框架 (SSRT-DETR),克服当前方法的局限性.
  • 通过提高匹配稳定性和自适应伪标签,在域移动下增强对象检测性能.

主要方法:

  • 利用一个平均的老师-学生架构与风格转移的图像,用于联合领域建模.
  • 引入域识别匹配 (DAM) 通过增加匈牙利匹配来稳定早期跨域培训.
  • 开发了类/场景适应性伪标签 (CAP),以根据类和场景特征动态调整值.

主要成果:

  • 在诸如Cityscapes→Foggy Cityscapes (mAP@0.5从51.0到54.3) 等基准上,SSRT-DETR显著提高了检测性能.
  • 在KITTI→Cityscapes和Sim10K→Cityscapes上实现了用于汽车检测的最先进的结果 (67.3 AP和64.9 AP).
  • 在罕见类别和恶劣天气中取得了持续的收益,验证了DAM和CAP的有效性,同时保持了实时效率.

结论:

  • 拟议的SSRT-DETR框架有效地解决了域自适应对象检测的挑战.
  • DAM和CAP模块是关键的创新,可以在多样化和具有挑战性的场景中实现强大的性能.
  • SSRT-DETR为实时,域自适应对象检测提供了一个有前途的解决方案.
关键词:
其他国家/地区 RT-DETRR域自适应对象检测对象检测域自适应对象检测半监督学习 半监督学习

更多相关视频

RBDT: A Computerized Task System based in Transposition for the Continuous Analysis of Relational Behavior Dynamics in Humans
11:09

RBDT: A Computerized Task System based in Transposition for the Continuous Analysis of Relational Behavior Dynamics in Humans

Published on: July 17, 2021

3.5K

相关实验视频

Last Updated: Mar 15, 2026

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

1.2K
RBDT: A Computerized Task System based in Transposition for the Continuous Analysis of Relational Behavior Dynamics in Humans
11:09

RBDT: A Computerized Task System based in Transposition for the Continuous Analysis of Relational Behavior Dynamics in Humans

Published on: July 17, 2021

3.5K