SaliencyI2PLoc: Saliency-guided image-point cloud localization using contrastive learning

Yuhao Li, Jianping Li, Zhen Dong, Yuan Wang, Bisheng Yang

Abstract: Highlights•We propose SaliencyI2PLoc, a Dual-Transformer for image-to-point cloud global localization.•Contrastive learning reduces explicit data mining preprocessing.•Saliency maps enhance local patch aggregation for better global features.•Feature relation consistency ensures robust cross-modality mapping.