Wenjun Li. "Scene-aware contrastive learning with multi-scale domain-invariant feature alignment for robust visual recognition under complex scenes." Scientific Reports, 2026. doi:10.1038/s41598-026-57271-1