Research map / Image generation / Autoregressive and latent generation
FIND: A Simple Yet Effective Baseline for Diffusion-Generated Image Detection
Jie Li, Yingying Feng, Chi Xie, Jie Hu, Lei Tan, Jiayi Ji
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 Gaussian noise to real images labeled synthetic (ablation) | FIND ViT-B/16 · GenImage (average of 8 generators) | accuracy: 80.58 -> 88.35 | Improved |
| add pure Gaussian noise images labeled synthetic (without noisy real images) (ablation) | FIND ViT-B/16 · GenImage (average of 8 generators) | accuracy: 80.58 -> 75.97 | Hurt |
| add pure Gaussian noise images on top of noisy real images (ablation) | FIND ViT-B/16 · GenImage (average of 8 generators) | accuracy: 88.35 -> 85.72 | Hurt |
| add noisy synthetic images labeled synthetic alone (ablation) | FIND ViT-B/16 · GenImage (average of 8 generators) | accuracy: 80.58 -> 74.97 | Hurt |
| add noisy synthetic images on top of noisy real images (ablation) | FIND ViT-B/16 · GenImage (average of 8 generators) | accuracy: 88.35 -> 84.50 | Hurt |
| switch backbone from ResNet50 to ViT-B/16 | FIND ViT-B/16 · GenImage (average of 8 generators) | accuracy: 80.13 -> 88.35 | Improved |
| replace reconstruction (LaRE2) with noise-augmented training, RN50 | FIND ViT-B/16 · GenImage (average of 8 generators) | accuracy: 79.07 -> 80.13 | Improved |
| remove reconstruction step versus DIRE | FIND ViT-B/16 · Inference speed | inference time: 151.9 ms -> 0.4 ms | Improved |
| remove reconstruction step versus LaRE2 | FIND ViT-B/16 · Inference speed | inference time: 50.4 ms -> 0.4 ms | Improved |
| increase noise magnitude epsilon beyond 20 (ablation) | FIND ViT-B/16 · GenImage SD V1.5 training subset | accuracy | No clear effect |
Closest papers by meaning
- FIRE: Robust Detection of Diffusion-Generated Images via Frequency-Guided Reconstruction Error (CVPR 2025 · similarity 0.89)
- Aligned Datasets Improve Detection of Latent Diffusion-Generated Images (ICLR 2025 · similarity 0.89)
- LaRE^2: Latent Reconstruction Error Based Method for Diffusion-Generated Image Detection (CVPR 2024 · similarity 0.88)
- Correcting Deviations from Normality: A Reformulated Diffusion Model for Multi-Class Unsupervised Anomaly Detection (CVPR 2025 · similarity 0.88)
- Beyond Generation: A Diffusion-based Low-level Feature Extractor for Detecting AI-generated Images (CVPR 2025 · similarity 0.88)
- AEROBLADE: Training-Free Detection of Latent Diffusion Images Using Autoencoder Reconstruction Error (CVPR 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.
