Continuous Adaptation of Multi-Camera Person Identification Models through Sparse Non-redundant Representative Selection.

•Continuous person identification with multi-sensor data is addressed.•An active learning set up involving human in the loop is used.•The main goal is to reduce human labeling effort but get good identification accuracy as more and more data becomes available.•A convex optimization based strategy progressively and judiciously chooses sparse and non-redundant set of samples for labeling.•Experiments on three publicly available benchmark datasets are performed to validate the proposed approach.