Scholarlink

<p>Detecting people and their attributes in natural conditions is one of the basic tasks of computer vision systems. Unfortunately, all known algorithms do not provide even the slightest protection of data privacy. In this paper, we fill this gap by proposing a system for detecting people while maintaining data privacy thanks to the learnable perceptual encryption of an image immediately after its acquisition and before detection. A novelty is our proposed hierachical image scrambling technique and modified YOLO detector, equipped with original input layers that enable effective detection in encrypted images. This is possible thanks to the use of the Lipschitz cost function and original attention modules. The experimental results using images from various everyday situations, obtained both in the visible spectrum and in the thermal band, prove the effectiveness of our method. Our code is available on GitHub.</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

Privacy-preserving people detection in the wild

  • 期刊论文

  • 发表日期：

    2025年4月15日

查看全文
  • Mateusz Knapik,
  • Bogusław Cyganek

摘要

Detecting people and their attributes in natural conditions is one of the basic tasks of computer vision systems. Unfortunately, all known algorithms do not provide even the slightest protection of data privacy. In this paper, we fill this gap by proposing a system for detecting people while maintaining data privacy thanks to the learnable perceptual encryption of an image immediately after its acquisition and before detection. A novelty is our proposed hierachical image scrambling technique and modified YOLO detector, equipped with original input layers that enable effective detection in encrypted images. This is possible thanks to the use of the Lipschitz cost function and original attention modules. The experimental results using images from various everyday situations, obtained both in the visible spectrum and in the thermal band, prove the effectiveness of our method. Our code is available on GitHub.