SegMAN-S

New training recipe boosts weather segmentation model generalization

Researchers have developed a new training methodology to improve the generalization capabilities of semantic segmentation models in adverse weather conditions. The study focused on five weather types: blur, darkness, sn…