Research map / Image generation / Autoregressive and latent generation
MarkDiffusion: An Open-Source Toolkit for Generative Watermarking of Latent Diffusion Models
Leyi Pan, Sheng Guan, Zheyu Fu, Luyang Si, Huan Wang, Zian Wang, Hanqian Li, Xuming Hu, Irwin King, Philip S. Yu, Aiwei Liu, Lijie Wen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Reimplement WIND in MarkDiffusion instead of official code | WIND · Image watermarking (SD v2.1, 200 prompts) | FID: 41.50 -> 37.91 (toolkit vs official) | Hurt |
| Reimplement Gaussian-Shading in MarkDiffusion | Gaussian-Shading · Image watermarking (SD v2.1, 200 prompts) | Gaussian Blur F1: 0.76 -> 0.82 (toolkit vs official) | Hurt |
| Reimplement Tree-Ring in MarkDiffusion | Tree-Ring · Image watermarking (SD v2.1, 200 prompts) | JPEG F1: 0.85 -> 0.88 (toolkit vs official) | Hurt |
| Reimplement Tree-Ring in MarkDiffusion | Tree-Ring · Image watermarking (SD v2.1, 200 prompts) | FID: 32.30 -> 33.15 (toolkit vs official) | Improved |
| Reimplement ROBIN in MarkDiffusion | ROBIN · Image watermarking (SD v2.1, 200 prompts) | FID: 23.15 -> 25.23 (toolkit vs official) | Improved |
| Reimplement ROBIN in MarkDiffusion | ROBIN · Image watermarking (SD v2.1, 200 prompts) | Gaussian Noise F1: 0.98 -> 1.00 (toolkit vs official) | Hurt |
| Reimplement SEAL in MarkDiffusion | SEAL · Image watermarking (SD v2.1, 200 prompts) | Detectability TPR@FPR=1%: 0.95 -> 0.98 (toolkit vs official) | Hurt |
| Reimplement Ring-ID in MarkDiffusion | Ring-ID · Image watermarking (SD v2.1, 200 prompts) | JPEG F1: 0.98 -> 1.00 (toolkit vs official) | Hurt |
| Reimplement PRC in MarkDiffusion | PRC · Image watermarking (SD v2.1, 200 prompts) | Gaussian Blur F1: 0.02 -> 0.00 (toolkit vs official) | No clear effect |
| Reimplement VideoShield in MarkDiffusion | VideoShield · Video watermarking (ModelScope, VBench prompts) | Avg of 5 quality metrics: 0.803 -> 0.804 (toolkit vs official) | No clear effect |
Closest papers by meaning
- WMAdapter: Adding WaterMark Control to Latent Diffusion Models (ICML 2025 · similarity 0.90)
- SleeperMark: Towards Robust Watermark against Fine-Tuning Text-to-image Diffusion Models (CVPR 2025 · similarity 0.90)
- Shallow Diffuse: Robust and Invisible Watermarking through Low-Dim Subspaces in Diffusion Models (NeurIPS 2025 · similarity 0.89)
- Harnessing Frequency Spectrum Insights for Image Copyright Protection Against Diffusion Models (CVPR 2025 · similarity 0.89)
- Black-Box Forgery Attacks on Semantic Watermarks for Diffusion Models (CVPR 2025 · similarity 0.88)
- AquaLoRA: Toward White-box Protection for Customized Stable Diffusion Models via Watermark LoRA (ICML 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
