Joint graph regularized dictionary learning and sparse ranking for multi-modal multi-shot person re-identification

Highlights：

• We propose to enforce the probe and gallery-based dual graph regularizer into the sparse ranking formulation, it can better capture the intrinsic geometric structure for multi-shot Re-ID.

• We propose to introduce a cross-modal regularizer for multi-modal person Re-ID, by enforcing the same person in different modal making similar contributions while reconstructing probe image from another modality.

• Comprehensive experiments on challenging benchmark datasets with both hand-crafted and deep features validate the superior performance of our model for multi-modal multi-shot person Re-ID.