Scholarlink

<p>With the advent of deep convolutional neural networks, human target detection has made significant progress. However, detecting humans through windows remains underexplored, primarily due to the absence of a dedicated dataset for window-based human detection. In this paper, we introduce a new dataset, WindowPerson, which establishes a baseline for detecting human targets at long range through building windows. Human detection near windows presents unique challenges, including complex scenes and cluttered occlusions, increasing the risk of false detection. Furthermore, we propose a gimbal control algorithm integrated with a Dahua network camera, which improves the human detection rate near windows by 50% compared to fixed cameras. We have open-sourced this data set on GitHub. You can access it at <a href="https://github.com/cyrusliu1984/WindowPerson-dataset">https://github.com/cyrusliu1984/WindowPerson-dataset</a>.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Window person detection dataset

  • 期刊论文

  • 发表日期：

    2025年7月4日

查看全文
  • Xiaochuan Liu,
  • Zhenyuan Yang,
  • Jianan Zhang,
  • Shiyue Xu,
  • Qin Yang,
  • Quanxiao Zhao

摘要

With the advent of deep convolutional neural networks, human target detection has made significant progress. However, detecting humans through windows remains underexplored, primarily due to the absence of a dedicated dataset for window-based human detection. In this paper, we introduce a new dataset, WindowPerson, which establishes a baseline for detecting human targets at long range through building windows. Human detection near windows presents unique challenges, including complex scenes and cluttered occlusions, increasing the risk of false detection. Furthermore, we propose a gimbal control algorithm integrated with a Dahua network camera, which improves the human detection rate near windows by 50% compared to fixed cameras. We have open-sourced this data set on GitHub. You can access it at https://github.com/cyrusliu1984/WindowPerson-dataset.