Lifelong Person Re-Identification with Backward-Compatibility

Minyoung Oh, Jae-Young Sim

Abstract: Lifelong person re-identification (LReID) assumes a practical scenario where the model is sequentially trained on continuously incoming datasets while alleviating the catastrophic forgetting in the old datasets. In such a case, not only the training datasets but the gallery images are also incrementally accumulated, that requires a huge amount storage space to store the gallery images as well as computational complexity to extract the features at the inference phase. In this paper, we address the above mentioned problem by incorporating the backward-compatibility to LReID based on the replay scheme. We attempt to train the model using the continuously incoming datasets while maintaining the model’s compatibility toward the previously trained old models without re-computing the features of the old gallery images. Specifically, we design the cross-model compatibility loss based on the contrastive learning with respect to the replay features across all the old datasets. Experimental results demonstrate that the proposed method achieves a significantly higher performance of the backward-compatibility compared with the existing LReID methods.