Researchers have developed a new metric called Remote Sensing Perceptual Distance (RSPD) to better evaluate the quality of generated remote sensing (RS) imagery. Existing metrics for natural images do not perform well with RS data, where visual plausibility can mask critical errors for downstream applications. RSPD utilizes layer-wise Aligned Attention Score (AAS) responses from RemoteCLIP, identifying that different RS artifacts are best detected at various network depths. This pairwise metric averages seven calibrated layer distances and demonstrates superior performance compared to simple cosine distance on realistic RS generation tasks, including super-resolution and SAR-to-optical translation. AI
IMPACT Enhances evaluation of AI-generated remote sensing imagery, crucial for applications relying on accurate geospatial data.
RANK_REASON The cluster contains a research paper detailing a new metric for image quality assessment in a specific domain. [lever_c_demoted from research: ic=1 ai=1.0]
- Aligned Attention Score
- alphaXiv
- arXiv
- CatalyzeX Code Finder for Papers
- CORE Recommender
- DagsHub
- Gotit.pub
- Hugging Face
- Influence Flower
- RemoteCLIP
- Remote Sensing Perceptual Distance
- ScienceCast
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →