Shilin-LU/VINE
[ICLR 2025] "Robust Watermarking Using Generative Priors Against Image Editing: From Benchmarking to Advances" (Official Implementation)
Leverages the SDXL-Turbo diffusion model as a generative prior to embed imperceptible watermarks while maintaining high image fidelity. Addresses watermark robustness against modern text-to-image editing by analyzing frequency characteristics and using blurring as a surrogate attack during training. Includes W-Bench, a comprehensive benchmark evaluating eleven watermarking methods across regeneration, global/local editing, and image-to-video generation tasks.
385 stars.
Stars
385
Forks
37
Language
Python
License
—
Category
Last pushed
Dec 01, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/Shilin-LU/VINE"
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...
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
bytedance/LatentUnfold
Implementation of paper: Flux Already Knows – Activating Subject-Driven Image Generation without Training