DiffIPE

DiffIPE will explore diffusion-based image processing and enhancement inside the Open-AISP pipeline.

The planned scope is to use generative priors for difficult restoration or enhancement stages while keeping camera consistency, controllability, and measurable fidelity as first-class constraints.

This module is currently a placeholder. Future documentation should cover:

  • task formulation
  • conditioning inputs
  • diffusion or distillation strategy
  • integration with RAW or linear RGB stages
  • evaluation protocol