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