Privacy Risks of Robot Vision: A User Study on Image Modalities and Resolution

  • 2025-05-12 18:16:12
  • Xuying Huang, Sicong Pan, Maren Bennewitz
  • 0

Abstract

User privacy is a crucial concern in robotic applications, especially whenmobile service robots are deployed in personal or sensitive environments.However, many robotic downstream tasks require the use of cameras, which mayraise privacy risks. To better understand user perceptions of privacy inrelation to visual data, we conducted a user study investigating how differentimage modalities and image resolutions affect users' privacy concerns. Theresults show that depth images are broadly viewed as privacy-safe, and asimilarly high proportion of respondents feel the same about semanticsegmentation images. Additionally, the majority of participants consider 32*32resolution RGB images to be almost sufficiently privacy-preserving, while mostbelieve that 16*16 resolution can fully guarantee privacy protection.

 

Quick Read (beta)

loading the full paper ...