CLEAR - Comparative Learning and Evaluation of AI and Traditional Denoisers

Comparative denoising study on large-scale UCF-101 video data with CV quality metrics.

CLEAR compared AI-based and traditional denoising methods under diverse noise conditions on large-scale video data.

CLEAR paper overview

Highlights

  • Led comparative analysis of denoising approaches for action-recognition workflows.
  • Built Python tooling with OpenCV/NumPy to inject multiple synthetic noise types.
  • Processed 13k+ videos using NVIDIA Tesla V100 compute for efficient batch evaluation.
  • Evaluated outputs with PSNR, SSIM, and VIF metrics.

Resources