Research map / Image generation / Autoregressive and latent generation
No Pixel Left Behind: A Detail-Preserving Architecture for Robust High-Resolution AI-Generated Image Detection
Lianrui Mu, Haoji Hu, Zou Xingze, Jianhong Bai, Jiaqi Hu
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 |
|---|---|---|---|
| cover all pixels with tiles instead of one center crop (ablation) | HiDA-Net FAM · GenImage | accuracy: 92.14 -> 96.10 | Improved |
| add Token-wise Forgery Localization (ablation) | FAM · GenImage | accuracy: 93.92 -> 94.36 | Improved |
| add JPEG Quality Factor Estimation (ablation) | FAM · GenImage | accuracy: 93.92 -> 94.73 | Improved |
| combine TFL and QFE (ablation) | FAM · GenImage | accuracy: 93.92 -> 96.10 | Improved |
| add QFE on in-distribution SD v1.4 JPEG robustness (ablation) | HiDA-Net · SD v1.4 JPEG robustness | accuracy | No clear effect |
| add SD v1.4 VAE paired fake images and patch swap (ablation) | HiDA-Net · GenImage | average accuracy: 96.1 -> 97.1 | Improved |
| add SD v1.4 VAE paired data and RPS (ablation) | HiDA-Net · DRCT test subsets | average accuracy: 94.8 -> 98.4 | Improved |
| replace resizing with full-coverage native-resolution tiles | detector · HiRes-50K | average accuracy: 80.33 vs 67.17 best baseline | Improved |
| full-coverage tiling vs TextureCrop on >5000 px images | detector · HiRes-50K >5000 | accuracy: 69.49 -> 69.84 | No clear effect |
Closest papers by meaning
- RealNet: Efficient and Unsupervised Detection of AI-Generated Images via Real-Only Representation Learning (AAAI 2026 · similarity 0.87)
- Towards High-Resolution Salient Object Detection (ICCV 2019 · similarity 0.86)
- Training-free Detection of AI-generated images via Cropping Robustness (NeurIPS 2025 · similarity 0.85)
- FerretNet: Efficient Synthetic Image Detection via Local Pixel Dependencies (NeurIPS 2025 · similarity 0.84)
- Is Artificial Intelligence Generated Image Detection a Solved Problem? (NeurIPS 2025 · similarity 0.84)
- Breaking Latent Prior Bias in Detectors for Generalizable AIGC Image Detection (NeurIPS 2025 · similarity 0.84)
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.
