Semantic-guided contrastive learning for SAR and optical image translation

This project is the official implementation code for the paper "SAR and Optical Image Conversion for Semantic Guided Comparative Learning" . To address the issues of feature homogenization and semantic ambiguity in cross modal transformation of remote sensing images, this project proposes a semantic guided contrastive learning framework. This framework introduces a semantic category consistency mechanism to accurately screen positive and negative samples in the contrastive learning space, and combines cyclic consistency loss and semantic segmentation loss for joint optimization.