Research map / Vision-language and multimodal / Cross-modal learning
Cross-modal Representation Learning for Diffusion-generated Image Detection
Tao Gong, Dayong Wang, Qi Chu, Bin Liu, Nenghai Yu
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 NPR modality to RGB (ablation) | SDID · GenImage | average accuracy: 86.7 -> 90.1 | Improved |
| add CMCL (ablation) | SDID · GenImage | average accuracy: 90.1 -> 95.4 | Improved |
| add CMMD (ablation) | SDID · GenImage | average accuracy: 95.4 -> 98.4 | Improved |
| add CMCL and CMMD with RGB-only inference (ablation) | SDID · GenImage | average accuracy: 86.7 -> 96.4 | Improved |
| use NPR instead of second RGB view for contrastive learning (ablation) | SDID · GenImage | average accuracy: 93.2 -> 98.4 | Improved |
| add CMMD to RGB & RGB input (ablation) | SDID · GenImage | average accuracy: 93.0 -> 93.2 | No clear effect |
| use high-frequency component instead of NPR (ablation) | SDID · GenImage | average accuracy: 98.4 -> 95.5 | Hurt |
| add CMCL and CMMD to CLIP RGB encoder (ablation) | SDID · GenImage | average accuracy: 89.3 -> 95.5 | Improved |
| SDID vs CoD on ADM test set | SDID · GenImage ADM | accuracy: 85.2 -> 95.9 | Improved |
| SDID vs DLFE on GenImage Midjourney and others | SDID · GenImage | average accuracy: 96.0 -> 98.4 | Improved |
| SDID on Co-Spy-Bench vs CO-SPY | SDID · Co-Spy-Bench | average accuracy: 87.1 -> 96.1 | Improved |
Closest papers by meaning
- On Learning Multi-Modal Forgery Representation for Diffusion Generated Video Detection (NeurIPS 2024 · similarity 0.89)
- FakeShield: Explainable Image Forgery Detection and Localization via Multi-modal Large Language Models (ICLR 2025 · similarity 0.88)
- Aligned Datasets Improve Detection of Latent Diffusion-Generated Images (ICLR 2025 · similarity 0.87)
- Spot the Fake: Large Multimodal Model-Based Synthetic Image Detection with Artifact Explanation (NeurIPS 2025 · similarity 0.87)
- FakeInversion: Learning to Detect Images from Unseen Text-to-Image Models by Inverting Stable Diffusion (CVPR 2024 · similarity 0.87)
- DiffusionFake: Enhancing Generalization in Deepfake Detection via Guided Stable Diffusion (NeurIPS 2024 · 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.
