Team Members

Amaro Truong

Nathan Thomas

Lyutfiya Yussupova

Emirhan Gencer

Project Overview

This research project explores the use of generative AI tools for restoring and enhancing degraded video and audio content. The work focuses on two key use cases: (1) improving the quality of dash cam footage by addressing common real-world issues such as glare, shadows, and obstructions like dirty windshields; and (2) restoring old or damaged films and television recordings, including enhancing resolution, removing visual artifacts, and cleaning up audio. The project begins with a survey of current state-of-the-art models and tools, followed by hands-on experimentation and the development of novel approaches tailored to these restoration scenarios. The ultimate goal is to create accessible, AI-powered solutions that preserve visual media and improve its clarity and impact for both archival and contemporary use.

Research Goals

  • Apply AI models for super-resolution and deblurring
  • Evaluate restoration fidelity using quantitative metrics
  • Create reusable restoration pipelines

Methods & Tools

  • GAN-based restoration (e.g.
  • ESRGAN)
  • Frame interpolation (e.g.
  • RIFE)
  • Python
  • FFmpeg
  • and OpenCV

Updates

Biweekly updates and demo videos will be posted here as the project progresses.