A Multimodal Anonymization Framework for MP4 Videos

Domain: Computer Vision + AIML

Award: Best Paper of the Session at IEEE International Conference on Emerging Smart Computing & Informatics (ESCI 2025), Pune.

Objective

Developed a multimodal anonymization framework for MP4 videos to ensure end-to-end privacy by integrating OCR for text redaction, GANs for facial anonymization, and audio processing for voice alteration.

Key Features

  • Optical Character Recognition (OCR) for text redaction
  • Generative Adversarial Networks (GANs) for facial anonymization
  • Audio processing for voice alteration
  • End-to-end privacy protection
  • Preservation of video usability
  • Real-time processing capabilities

Technologies Used

  • Python
  • OpenCV
  • GANs (Generative Adversarial Networks)
  • OCR (Optical Character Recognition)
  • Audio Processing Libraries
  • Computer Vision

Applications

  • Healthcare - Patient privacy protection
  • Legal - Evidence anonymization
  • Social Media - Content privacy
  • Privacy-focused ML applications

Impact

  • Comprehensive privacy protection for video content
  • Maintains video usability while protecting sensitive data
  • Effective for multiple industry applications
  • Recognized with Best Paper award at IEEE conference
  • Advancement in privacy-preserving AI technology