Researchers have developed CV-SSMNet, a novel physics-aware complex-valued state-space network designed for polarimetric synthetic aperture radar (PolSAR) image classification. This model addresses limitations in existing networks by incorporating physical scattering mechanisms directly into deep feature evolution, rather than treating them as shallow auxiliary inputs. CV-SSMNet utilizes a complex-valued state-space model to capture long-range spatial dependencies while preserving amplitude-phase information, and employs scattering priors as modulation signals to adaptively recalibrate representations. Experiments on L-band and P-band datasets show that CV-SSMNet achieves competitive accuracy, enhanced regional consistency, and improved boundary preservation. AI
IMPACT This model could improve the accuracy and consistency of GeoAI applications that rely on detailed image classification.
RANK_REASON The cluster describes a new academic paper detailing a novel model for image classification.
Read on Hugging Face Daily Papers →
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
