Unsupervised Noisy Tracklet Person Re-identification
This work addresses scalability and cost issues in person re-identification for large-scale deployments by enabling unsupervised learning from unlabeled tracklets, though it is incremental as it builds on existing unsupervised methods.
The authors tackled the problem of training person re-identification models without labeled data by proposing a selective tracklet learning approach that is robust to noisy tracklets, achieving significant improvements over state-of-the-art unsupervised and one-shot learning methods on three large benchmarks.
Existing person re-identification (re-id) methods mostly rely on supervised model learning from a large set of person identity labelled training data per domain. This limits their scalability and usability in large scale deployments. In this work, we present a novel selective tracklet learning (STL) approach that can train discriminative person re-id models from unlabelled tracklet data in an unsupervised manner. This avoids the tedious and costly process of exhaustively labelling person image/tracklet true matching pairs across camera views. Importantly, our method is particularly more robust against arbitrary noisy data of raw tracklets therefore scalable to learning discriminative models from unconstrained tracking data. This differs from a handful of existing alternative methods that often assume the existence of true matches and balanced tracklet samples per identity class. This is achieved by formulating a data adaptive image-to-tracklet selective matching loss function explored in a multi-camera multi-task deep learning model structure. Extensive comparative experiments demonstrate that the proposed STL model surpasses significantly the state-of-the-art unsupervised learning and one-shot learning re-id methods on three large tracklet person re-id benchmarks.