Research Explorer

Project

User guided factorized filtering out of privacy-sensitive attributes in video data
Code
01D09322
Duration
01 October 2022 → 31 August 2026
Funding
Regional and community funding: Special Research Fund
Research disciplines
  • Natural sciences
    • Computer vision
Keywords
Privacy preserving machine learning disentangled representation learning deep generative modelling
 
Project description
 

This project researches the design of filters that can remove sensitive information from video without losing the ability to perform allowed tasks. We will do this by learning a factorized and disentangled representation through generative modelling. By modifying select factors of this representation we aim to remove all mutual information between the image and the sensitive
information.