Research map / Image generation / Fast diffusion sampling
D3F-IR: Dual-Domain Deterministic Flow Matching for Visible-to-Infrared Translation
Peiyi Zeng, Diedong Feng, Zhen Liu, Zhenming Peng, Bing Zeng, Shuaicheng Liu
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 the pixel-domain flow branch to semantic-only flow (ablation) | D3F-IR · FLIR | FID: 74.542 -> 66.716 | Improved |
| add the semantic-domain flow branch to pixel-only flow (ablation) | D3F-IR · FLIR | FID: 163.088 -> 66.716 | Improved |
| add the semantic branch to pixel-only flow (ablation) | D3F-IR · KAIST | PSNR: 22.079 -> 24.938 | Improved |
| add the semantic flow objective to pixel-only objective (ablation) | D3F-IR · FLIR | FID: 103.299 -> 66.716 | Improved |
| add pixel flow objective to semantic-only objective (ablation) | D3F-IR · KAIST | PSNR: 24.692 -> 24.938 | Improved |
| dual-domain flow versus single-domain deterministic flow (ablation) | D3F-IR · FLIR | PSNR: 18.938 -> 20.094 | Improved |
| deterministic flow versus standard diffusion (ablation) | single-domain generation · FLIR | FID: 158.517 -> 74.542 | Improved |
| patch-wise modulation instead of cross-attention in SEAM (ablation) | D3F-IR · FLIR | FID: 114.071 -> 66.716 | Improved |
| patch-wise modulation instead of concatenation in SEAM (ablation) | D3F-IR · KAIST | FID: 85.622 -> 75.909 | Improved |
| add external infrared pre-training before fine-tuning | D3F-IR · KAIST | FID: 75.909 -> 40.312 | Improved |
| add external infrared pre-training | D3F-IR · FLIR | PSNR: 20.094 -> 19.581 | Hurt |
| pre-training on M3FD | D3F-IR · M3FD | FID: 103.264 -> 65.750 | Improved |
Closest papers by meaning
- CoVisIT: Cross-modal Prior Guided Diffusion Model for Visible-to-Infrared Image Translation (NeurIPS 2026 · similarity 0.89)
- TherA: Thermal-Aware Visual-Language Prompting for Controllable RGB-to-Thermal Infrared Translation (CVPR 2026 · similarity 0.89)
- ThermalGen: Style-Disentangled Flow-Based Generative Models for RGB-to-Thermal Image Translation (NeurIPS 2025 · similarity 0.88)
- IRCasDiff: Two-Stage Cascaded Diffusion for Compound Infrared Face Reconstruction (NeurIPS 2026 · similarity 0.86)
- Beyond Strict Pairing: Arbitrarily Paired Training for High-Performance Infrared and Visible Image Fusion (CVPR 2026 · similarity 0.86)
- SynthRGB-T: Language-Vision Guided Image Translation for Diversity Synthesis (CVPR 2026 · 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.
