Skip to content

Image generation explainability

Under construction

Image-generation explainability is not yet available as a supported XWhy workflow. The repository currently exposes an early Pix2PixExplainer interface, but its explain() method is not implemented.

This section is the public entry point for explainability methods applied to systems that create or transform images. Pix2Pix is treated as one model-family example within this broader capability, not as the name of the capability itself.

Planned coverage

Image-generation documentation is expected to cover:

  • text-to-image generation;
  • image-to-image generation;
  • instruction-guided image editing;
  • inpainting and region replacement;
  • conditional generation;
  • attribution to prompts, source-image regions, masks, and conditioning inputs;
  • output similarity and perceptual-distance measures;
  • explanation stability, faithfulness, and safety limitations.

Subsections

Current implementation status

The current code contains Pix2PixExplainer as an early interface. It should remain described as an implementation prototype until its behaviour, accepted inputs, perturbation strategy, result type, tests, and examples are complete.

View the current Pix2Pix API interface