Research map / Image generation / Image restoration
OneHSI: A Unified Hyperspectral Foundation Model with Physical Consistency
Yufei WEN, Jingdan KANG, SHUXIN ZHONG, Yuting Zhang, Yutong Feng, Jintai Chen, Kaishun Wu
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 statistical (FiLM) modulation after spectral alignment (ablation) | OneHSI · OHID-1 | OA: 91.65 -> 92.50 | Improved |
| use learnable wavelength reparameterization instead of CNN projection (ablation) | OneHSI · OHID-1 | OA: 90.27 -> 92.50 | Improved |
| use learnable wavelength reparameterization instead of interpolation (ablation) | OneHSI · OHID-1 | OA: 85.10 -> 92.50 | Improved |
| add domain-adversarial loss (ablation) | OneHSI · OHID-1 | OA: 89.73 -> 92.50 | Improved |
| add expert balancing loss (ablation) | OneHSI · OHID-1 | OA: 90.68 -> 92.50 | Improved |
| use scale-aware SEAT routing instead of generic MoE (top-2) (ablation) | OneHSI · HeiPorSPECTRAL | OA: 94.55 -> 96.81 | Improved |
| generic MoE instead of dense network (ablation) | OneHSI · OHID-1 | OA: 89.85 -> 90.80 | Improved |
| top-2 expert routing instead of top-1 (ablation) | SEAT · OHID-1 | OA: 91.95 -> 92.50 | Improved |
| pretrain jointly on all four domains instead of sequentially / dropping one (ablation) | OneHSI · downstream OA | OA | Improved |
| OneHSI vs HyperFree with identical pretraining data (frozen encoder) (ablation) | OneHSI · OHID-1 | OA: 86.14 -> 90.65 | Improved |
| OneHSI on NPU detection vs SS-Mamba | OneHSI · NPU | mAP: 78.33 vs 77.68 | No clear effect |
| fewer shots (1-shot) with OneHSI vs HyperFree | OneHSI · OHID-1 1-shot | OA: 78.41 -> 80.76 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- HyperSIGMA — HyperSIGMA: Hyperspectral Intelligence Comprehension Foundation Model (TPAMI 2025) · better on 10 of 10 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Bridging Local–Global Dissonance: Learning from Compressive Measurements for Hyperspectral Reconstruction (ICML 2026 · similarity 0.89)
- PhySpec: Physically Consistent Spectral Reconstruction via Orthogonal Subspace Decomposition and Self-Supervised Meta-Auxiliary Learning (ICML 2025 · similarity 0.87)
- EigenSR: Eigenimage-Bridged Pre-Trained RGB Learners for Single Hyperspectral Image Super-Resolution (AAAI 2025 · similarity 0.87)
- LESSViT: Robust Hyperspectral Representation Learning under Spectral Configuration Shift (NeurIPS 2026 · similarity 0.87)
- Hyperspectral Image Reconstruction via Combinatorial Embedding of Cross-Channel Spatio-Spectral Clues (AAAI 2024 · similarity 0.86)
- FRN: Fractal-Based Recursive Spectral Reconstruction Network (NeurIPS 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.
