CrossCD-Eval Framework Enhances Cross-Domain Remote Sensing Change Detection

This announcement introduces CrossCD-Eval, a unified framework designed to evaluate the generalization of deep learning models for change detection in remote sensing imagery across different domains. The framework addresses challenges caused by distribution shifts from geographic and sensor differences and proposes domain-aware feature normalization to reduce domain mismatch. Five recent models were evaluated using multiple datasets to test performance under various domain adaptation settings.

Originally published by gnews