LG-Diff: Learning to follow local class-regional guidance for nearshore image cross-modality high-quality translation

Jiangang Ding, Yiquan Du, Wei Li, Lili Pei, Ningning Cui

Abstract: Highlights•Nearshore cross-modal translation is formulated as a denoising diffusion process.•We develop a two-stage pipeline guided by local class-regional information.•Decoupling high- and low-frequency information in the feature extraction.•A cross-modal benchmark for nearshore image translation task is proposed.