MPCCL: Multiview predictive coding with contrastive learning for person re-identification

Highlights：

• We propose the first re-ID network that implicitly encodes the fine-grained semantic information of input data into the embedding space and learns semantic-aware representation via representation reconstruction.

• We propose multi-view predictive coding to align different representations of the same person to maintain intra-class similarity.

• Representation reconstruction and contrastive learning jointly supervise the representation learning process, thus obtaining fine-grained semantic information and appearance-free representations.

• The proposed method achieves the state-of-the-art performance on several benchmark datasets