yunqing-me/WatermarkDM
Code of the paper: A Recipe for Watermarking Diffusion Models
Implements watermarking for both unconditional diffusion models (via binary bit-string embedding with decoder recovery) and text-to-image models like Stable Diffusion (via trigger-based image-text pair injection). Uses a two-stage approach: training watermark encoder/decoder networks on PyTorch, then embedding watermarks into training data before fine-tuning diffusion models with distributed training across multiple GPUs on datasets like CIFAR-10, FFHQ, and AFHQv2.
155 stars. No commits in the last 6 months.
Stars
155
Forks
11
Language
Jupyter Notebook
License
MIT
Category
Last pushed
Nov 13, 2024
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/yunqing-me/WatermarkDM"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
OPTML-Group/Unlearn-Saliency
[ICLR24 (Spotlight)] "SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in...
Shilin-LU/VINE
[ICLR 2025] "Robust Watermarking Using Generative Priors Against Image Editing: From...
WindVChen/DiffAttack
An unrestricted attack based on diffusion models that can achieve both good transferability and...
koninik/DiffusionPen
Official PyTorch Implementation of "DiffusionPen: Towards Controlling the Style of Handwritten...
Wuyxin/DISC
(ICML 2023) Discover and Cure: Concept-aware Mitigation of Spurious Correlation