About this document
Unsupervised Person Re-ID Noise Purification by bkdaddin is a document available to read on EtoBox.
This article proposes a method for unsupervised person re-identification that aims to purify feature and label noise. It introduces two purification modules: 1) A multi-view feature module that enriches the global feature representation with additional local view features. 2) A knowledge distillation module that trains a teacher model on noisy pseudo labels to guide learning and reduce interference from label noise. The purification modules are integrated into an existing cluster contrast learning framework
- Author
- bkdaddin
- Language
- EN