Unsupervised person re-identification via simultaneous clustering and mask prediction

作者：

Highlights：

• We design mask prediction as a pretext task for unsupervised re-ID by learning visual consistency from still images and temporal consistency during training process.

• We optimize the model by grouping the two encoded views into the same cluster, thus enhancing the visual consistency between views.

• The proposed clustering network can separate images into semantic clusters automatically.

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

摘要

•We design mask prediction as a pretext task for unsupervised re-ID by learning visual consistency from still images and temporal consistency during training process.•We optimize the model by grouping the two encoded views into the same cluster, thus enhancing the visual consistency between views.•The proposed clustering network can separate images into semantic clusters automatically.•The proposed method achieves the state-of-the-art performance on several benchmark datasets.

论文关键词：Person re-identification,Domain adaptation,Unsupervised clustering,Mask prediction,Semantic cluster

论文评审过程：Received 20 August 2021, Revised 7 January 2022, Accepted 31 January 2022, Available online 1 February 2022, Version of Record 13 February 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108568