Skip to content

diffusion-e2e-ft

Fine-tune image-conditioned diffusion models end to end

Visit official site

What diffusion-e2e-ft is

diffusion-e2e-ft is described as an open-source tool for fine-tuning image-conditioned diffusion models so pretrained models perform better on specific tasks.

What you can do with it

  • Tune several conditional models
  • Load pretrained checkpoints
  • Choose noise types and time steps

Application scenarios

  • A depth map becomes more accurate
  • Surface normals improve in hard scenes
  • An analysis pipeline gains accuracy

Getting started

  • Clone the repository locally
  • Install the required dependencies
  • Select a checkpoint for fine-tuning
  • Configure inference parameters

Pros & cons

✓ What we like

  • Detailed training and inference instructions are provided
  • Half-precision operation reduces resource use

! What to watch out for

  • Fine-tuning starts from a chosen checkpoint
  • Inference parameters must be configured by hand

FAQ

Which tasks are named?

Depth estimation and normal estimation are named in the material.

Which checkpoints are listed?

Marigold and GeoWizard checkpoints are listed.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review