Abstract: Highlights•Multi-view dataset proposed to facilitate view-based person ReID research.•Grouping mechanism and cross-modal fusion enable efficient person ReID.•It achieves competitive performance for occluded and holistic person ReID.
Loading