Research map / Vision-language and multimodal / Cross-modal learning
Rethinking Cross-Spectral Image Generation via Shared-Specific Representation
Haining Wang, Na Li, Huijie Zhao, Yifan Da, Yan Wen, Yi Su, Yuqiang Fang
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 semantic refinement (adversarial structure discriminator) to the encoders (ablation) | SHASP · AVIID | FID: 45.39 -> 41.08 | Improved |
| add semantic refinement (ablation) | SHASP · AVIID | SSIM: 0.80 -> 0.85 | Improved |
| add semantic refinement (ablation) | SHASP · DroneCoast | FID: 136.95 -> 125.28 | Improved |
| add structure content encoder to CycleGAN baseline (ablation) | CycleGAN · AVIID | FID: 58.87 -> 50.76 | Improved |
| add spectral feature encoder to CycleGAN baseline (ablation) | CycleGAN · AVIID | FID: 58.87 -> 54.23 | Improved |
| add both structure and spectral encoders to CycleGAN (ablation) | CycleGAN · AVIID | FID: 58.87 -> 45.39 | Improved |
| add both encoders to CycleGAN (ablation) | CycleGAN · DroneCoast | FID: 170.90 -> 136.95 | Improved |
| enlarge training data 4x with label-preserving augmentation (ablation) | SHASP · AVIID | FID: 41.08 -> 40.15 | Improved |
| double semantic loss weight alpha from 1 to 2 (ablation) | SHASP · AVIID | FID: 43.77 -> 41.08 | Improved |
| double cycle loss weight beta to 2 (ablation) | SHASP · AVIID | FID: 43.77 -> 42.29 | Improved |
| double adversarial weight delta to 2 (ablation) | SHASP · AVIID | SSIM: 0.846 -> 0.835 | Hurt |
| train detector on SHASP-generated IR instead of real VIS images | YOLOv9 · AVIID and DroneCoast detection | AP50 | Improved |
Closest papers by meaning
- Cross-Spectral Face Hallucination via Disentangling Independent Factors (CVPR 2020 · similarity 0.87)
- Cross-Spectral Gaussian Splatting with Spatial Occupancy Consistency (AAAI 2025 · similarity 0.86)
- DSGCR: Decomposed Spectral Geometry-Aware Cross-Modal Semantic Representation for 3D Visual Grounding (ICML 2026 · similarity 0.84)
- Spectral Compressive Imaging via Chromaticity-Intensity Decomposition (NeurIPS 2025 · similarity 0.83)
- Monocular Identity-Conditioned Facial Reflectance Reconstruction (CVPR 2024 · similarity 0.83)
- Hyperspectral Image Reconstruction via Combinatorial Embedding of Cross-Channel Spatio-Spectral Clues (AAAI 2024 · similarity 0.82)
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.
