Dynamic Label Graph Matching for Unsupervised Video Re-identification

By Mang Ye, Andy Jinhua Ma, Liang Zheng, Jiawei Li, Pong Chi Yuen. 2017

194 citations recorded by the source corpus.

Label estimation is an important component in an unsupervised person re-identification (re-ID) system. This paper focuses on cross-camera label estimation, which can be subsequently used in feature learning to learn robust re-ID models. Specifically, we propose to construct a graph for samples in each camera, and then graph matching scheme is introduced for cross-camera labeling association. While labels directly output from existing graph matching methods may be noisy and inaccurate due to significant cross-camera variations, this paper propose a dynamic graph matching (DGM) method. DGM iteratively updates the image graph and the label estimation process by learning a better feature space with intermediate estimated labels. DGM is advantageous in two aspects: 1) the accuracy of estimated labels is improved significantly with the iterations; 2) DGM is robust to noisy initial training data. Extensive experiments conducted on three benchmarks including the large-scale MARS dataset show that DGM yields competitive performance to fully supervised baselines, and outperforms competing unsupervised learning methods.1

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Unsupervised Embedding Learning via Invariant and Spreading Instance Feature (2019, neutral): “nearest neighbor search [46, 48, 50].”
Image-Image Domain Adaptation with Preserved Self-Similarity and Domain-Dissimilarity for Person Re-identification (2017, methodology): “[47] use graph matching for cross-camera label estimation.”
A Bottom-Up Clustering Approach to Unsupervised Person Re-Identification (2019, methodology): “There are also some recent works (Liu, Wang, and Lu 2017; Ye et al. 2017; Ye, Lan, and Yuen 2018) focusing on the un-supervised video-based re-ID.”
A Pose-Sensitive Embedding for Person Re-identification with Expanded Cross Neighborhood Re-ranking (2017, unknown): “6 DGM+IDE [38] ICCV17 - - - - 46.”
Deep Learning for Person Re-Identification: A Survey and Outlook (2020, background): “The success of deep learning on these closed-world datasets also motivates the shift focus to more challenging scenarios, i.e ., large data size [140] or unsupervised learning [127].”
Robust Anchor Embedding for Unsupervised Video Person re-IDentification in the Wild (2018, methodology): “Compared with DGM [47], RACE is more robust to lower positive ratios, while DGM drops quickly with low positive ratios.”
Exploit the Unknown Gradually: One-Shot Video-Based Person Re-identification by Stepwise Learning (2018, methodology): “The key challenge for the one-shot video-based person re-ID is the label estimation for the abundant unlabeled tracklets [7, 34].”
Weakly Supervised Segmentation with Maximum Bipartite Graph Matching (2020, methodology): “[61] utilize an on-line and off-line transfer matching to person Re-ID task.”
Recognition From Web Data: A Progressive Filtering Approach (2018, unknown): “vision [22]–[24], etc.”
Differentiation of Blackbox Combinatorial Solvers (2019, background): “Examples include multi-object tracking [41], semantic segmentation [7], multi-person pose estimation [35, 44], stereo matching [23] and person re-identification [52].”
Hierarchical Discriminative Learning for Visible Thermal Person Re-Identification (2018, methodology): “Under this visible domain to model the large cross-camera variations, encouraging results have been achieved (Ye et al. 2017) by 1) learning or designing representative features for visual appearance of persons, and 2) learning proper similarity measurements on the visual feature vectors.”