Research map / Vision-language and multimodal / CLIP and zero-shot
G-ZAP: A Generalizable Zero-Shot Framework for Arbitrary-Scale Pansharpening
Zhiqi Yang, Shan Yin, Jingze Liang, Liang-Jian Deng
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 |
|---|---|---|---|
| remove-from-full: add full-resolution spectral consistency loss L0 (ablation) | G-ZAP · WV3 | HQNR: 0.9276 -> 0.9585 | Improved |
| add L0 (ablation) | G-ZAP · WV2 (cross-sensor) | HQNR: 0.8781 -> 0.9226 | Improved |
| add reduced-resolution supervision L1 (ablation) | G-ZAP · WV3 | HQNR: 0.9377 -> 0.9585 | Improved |
| add reduced-resolution supervision L1 (ablation) | G-ZAP · WV2 (cross-sensor) | HQNR: 0.8970 -> 0.9226 | Improved |
| add multi-scale supervision L2 (ablation) | G-ZAP · WV3 | HQNR: 0.9511 -> 0.9585 | Improved |
| add multi-scale supervision L2 (ablation) | G-ZAP · WV2 (cross-sensor) | Dlambda: 0.0264 -> 0.0304 | Hurt |
| reuse weights trained on one pair instead of per-image training | G-ZAP · WV3 | HQNR: 0.9585 -> 0.9574 | No clear effect |
| reuse weights from WV3 pair on WV2 | G-ZAP · WV2 (cross-sensor) | HQNR: 0.9226 -> 0.9367 | Improved |
| weight reuse instead of per-image training | G-ZAP · WV3 running time | time (s): 79.03 -> 4.23 | Improved |
| query G-ZAP directly at target scale instead of x1 output + bicubic | G-ZAP · Reduced-resolution WV3, scale 6.7 | Q2n: 0.4107 -> 0.4926 | Improved |
| replace INRConv backbone with FusionNet (convolutional) (ablation) | G-ZAP · WV2 weight reuse | performance | Hurt |
Closest papers by meaning
- Cross-Scale Pansharpening via ScaleFormer and the PanScale Benchmark (CVPR 2026 · similarity 0.88)
- Unsupervised Pan-Sharpening via Mutually Guided Detail Restoration (AAAI 2024 · similarity 0.86)
- PanAdapter: Two-Stage Fine-Tuning with Spatial-Spectral Priors Injecting for Pansharpening (AAAI 2025 · similarity 0.86)
- Hyperspectral Pansharpening via Diffusion Models with Iteratively Zero-Shot Guidance (CVPR 2025 · similarity 0.86)
- Wavelet-Assisted Multi-Frequency Attention Network for Pansharpening (AAAI 2025 · similarity 0.86)
- Zero-Shot Scene Reconstruction from Single Images with Deep Prior Assembly (NeurIPS 2024 · 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.
