Safeguard with Human Feedback Inversion

September 18, 2024 · View on GitHub

Official implementation of "Safeguard Text-to-Image Diffusion Models with Human Feedback Inversion" (ECCV 2024)

WIP 🚧

We are currently finalizing the human annotation data and UI. The code and data will be available before the conference.

See https://github.com/nannullna/safe-diffusion for Safe self-Distillation Diffusion (SDD).
See https://huggingface.co/docs/diffusers/training/text_inversion and https://github.com/huggingface/diffusers/blob/main/examples/textual_inversion/textual_inversion.py for Textual Inversion.