Skip to content

SHMT

Transfer makeup between faces with a latent diffusion model

Visit official site

What SHMT is

The technique is described as a self-supervised hierarchical makeup transfer method using a latent diffusion model, accepted at NeurIPS 2024.

What you can do with it

  • Train without annotated datasets
  • Transfer looks across hierarchical levels
  • Combine images, segmentation maps and depth maps

Getting started

  • Set up Python and the required libraries
  • Clone the project code from GitHub
  • Place the pre-trained weights and edit the config

Pros & cons

✓ What we like

  • Pre-trained models are available for quick tasks
  • Model configurations can be adjusted by level

! What to watch out for

  • Setup asks for a Python environment and dependencies
  • Pre-trained models must be downloaded separately

FAQ

Does it need labelled data?

The material states training is self-supervised and needs no annotations.

Where was it published?

The material states it was accepted at NeurIPS 2024.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review