Research map / Detection and segmentation / CNN features
Decision-Driven Orthogonal Learning with Complementary Feature Mining for Robust Synthetic Image Detection
Kai Li, Wei Wang, Linchao Zhang, Siying Zhu, Wenqi Ren
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 decision-driven orthogonal constraint (ablation) | Ours (ViT + conv branch) · ProGAN 2-class, quality-aware | Mean accuracy (%): 72.3 -> 75.4 | Improved |
| add adaptive global-local modulator (ablation) | Ours without orthogonal constraint · ProGAN 2-class, quality-aware | Mean accuracy (%): 70.9 -> 72.3 | Improved |
| add bidirectional update strategy (ablation) | Plain backbone · ProGAN 2-class, quality-aware | Mean accuracy (%): 63.5 -> 70.9 | Improved |
| replace gradient reversal compression removal with orthogonal constraint | ODDN · 2-class quality-aware | Mean accuracy (%): 71.4 -> 75.4 | Improved |
| same method in unknown-compression setting | ODDN · 2-class quality-agnostic | Mean accuracy (%): 70.7 -> 74.5 | Improved |
| same method with 4-class training data | ODDN · 4-class quality-aware | Mean accuracy (%): 72.6 -> 76.5 | Improved |
| same method with 4-class training, unknown compression | ODDN · 4-class quality-agnostic | Mean accuracy (%): 72.1 -> 75.5 | Improved |
| proposed method vs ODDN on AttGAN | ODDN · 2-class quality-aware | Accuracy (%): 71.2 -> 54.8 | Hurt |
| proposed method vs ODDN on BEGAN | ODDN · 2-class quality-aware | Accuracy (%): 42.3 -> 64.9 | Improved |
| proposed method vs ODDN on StarGAN | ODDN · 2-class quality-aware | Accuracy (%): 77.3 -> 92.3 | Improved |
| proposed method vs ODDN on RelGAN | ODDN · 2-class quality-aware | Accuracy (%): 72.5 -> 58.7 | Hurt |
| add DDOC and modules (logit separation) | baseline detector · ProGAN and GauGAN test | logit distribution overlap | Improved |
Closest papers by meaning
- Learning Discriminative Noise Guidance for Image Forgery Detection and Localization (AAAI 2024 · similarity 0.86)
- Forgery-aware Adaptive Transformer for Generalizable Synthetic Image Detection (CVPR 2024 · similarity 0.85)
- Identity-Aware Vision-Language Model for Explainable Face Forgery Detection (AAAI 2026 · similarity 0.85)
- Stay-Positive: A Case for Ignoring Real Image Features in Fake Image Detection (ICML 2025 · similarity 0.85)
- A Bias-Free Training Paradigm for More General AI-generated Image Detection (CVPR 2025 · similarity 0.85)
- A Lottery Ticket Hypothesis Approach with Sparse Fine-tuning and MAE for Image Forgery Detection and Localization (AAAI 2025 · similarity 0.85)
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.
