Research map / Image generation / Autoregressive and latent generation
FingerprintNet: Synthesized Fingerprints for Generated Image Detection
Yonghyun Jeong, Doyeon Kim, Youngmin Ro, Pyounggeon Kim, Jongwon Choi
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 |
|---|---|---|---|
| add random layer selection to the fingerprint generator (ablation) | FingerprintNet · Cross-model mean | accuracy: 60.0 -> 82.6 | Improved |
| add multi-kernel deconvolution layer (ablation) | FingerprintNet · Cross-model mean | accuracy: 71.3 -> 82.6 | Improved |
| add feature blender (ablation) | FingerprintNet · Cross-model mean | accuracy: 74.9 -> 82.6 | Improved |
| add feature similarity loss (ablation) | FingerprintNet · Cross-model mean | accuracy: 74.8 -> 82.6 | Improved |
| add mixed-batch training with all-fake-if-any-fake labels (ablation) | FingerprintNet · Cross-model mean | accuracy: 74.6 -> 82.6 | Improved |
| add FFT magnitude spectrum input (ablation) | FingerprintNet · Cross-model mean | average precision: 82.9 -> 94.0 | Improved |
| self-supervised real-only training vs supervised on 20 ProGAN classes | FingerprintNet vs Frank · Cross-model mean | accuracy: 77.1 -> 82.6 | Improved |
| self-supervised vs supervised detector on non-GAN generators | FingerprintNet vs Frank · DDPM, ILVR, NVAE, FICGAN (mean) | accuracy: 87.5 vs 86.7 | No clear effect |
| FingerprintNet vs SelfDetector on cross-category CycleGAN | FingerprintNet · CycleGAN cross-category (mean of 6 classes) | accuracy: 86.9 -> 93.8 | Improved |
Closest papers by meaning
- One for All: Synthesis-Free Fingerprint Learning for Attribution of In-the-Wild Synthetic Images (AAAI 2026 · similarity 0.93)
- ManiFPT: Defining and Analyzing Fingerprints of Generative Models (CVPR 2024 · similarity 0.91)
- PrintsGAN: Synthetic Fingerprint Generator (TPAMI 2022 · similarity 0.90)
- WOUAF: Weight Modulation for User Attribution and Fingerprinting in Text-to-Image Diffusion Models (CVPR 2024 · similarity 0.87)
- OmniMark: Efficient and Scalable Latent Diffusion Model Fingerprinting (AAAI 2025 · similarity 0.87)
- Scalable Fingerprinting of Large Language Models (NeurIPS 2025 · similarity 0.86)
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.
