Unsupervised Person Re-identification

By Hehe Fan, Liang Zheng, Chenggang Clarence Yan, Yi Yang. ACM Trans. Multim. Comput. Commun. Appl., 2017

508 citations recorded by the source corpus.

The superiority of deeply learned pedestrian representations has been reported in very recent literature of person re-identification (re-ID). In this article, we consider the more pragmatic issue of learning a deep feature with no or only a few labels. We propose a progressive unsupervised learning (PUL) method to transfer pretrained deep representations to unseen domains. Our method is easy to implement and can be viewed as an effective baseline for unsupervised re-ID feature learning. Specifically, PUL iterates between (1) pedestrian clustering and (2) fine-tuning of the convolutional neural network (CNN) to improve the initialization model trained on the irrelevant labeled dataset. Since the clustering results can be very noisy, we add a selection operation between the clustering and fine-tuning. At the beginning, when the model is weak, CNN is fine-tuned on a small amount of reliable examples that locate near to cluster centroids in the feature space. As the model becomes stronger, in subsequent iterations, more images are being adaptively selected as CNN training samples. Progressively, pedestrian clustering and the CNN model are improved simultaneously until algorithm convergence. This process is naturally formulated as self-paced learning. We then point out promising directions that may lead to further improvement. Extensive experiments on three large-scale re-ID datasets demonstrate that PUL outputs discriminative features that improve the re-ID accuracy. Our code has been released at https://github.com/hehefan/Unsupervised-Person-Re-identification-Clustering-and-Fine-tuning.

Subfields: Biomedical Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/c7fff0d0a6312965b269c6180b2112babd40564c

How researchers see this paper

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

In Biomedical Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Deep Learning for Person Re-Identification: A Survey and Outlook (2020, methodology): “These methods include PUL [208], CAMEL [206], PTGAN [118], TJ-AIDL [109], HHL [215], DAS [122], MAR [210], ENC [105], ATNet [216], PAUL [157], SBGAN [217], UCDA [219], CASC [211], PDA [119], CR-GAN [218], PAST [220], SSG [213].”
Style Normalization and Restitution for Generalizable Person Re-Identification (2020, methodology): “UDAbased ReID methods can be roughly divided into three categories: style transfer [3, 44, 26], attribute recognition [43, 49, 33], and target-domain pseudo label estimation [5, 37, 58, 40, 52, 50].”
Unsupervised Person Re-Identification via Multi-Label Classification (2020, methodology): “We compare two types of methods, including unsupervised learning methods: LOMO [18], BOW [42], BUC [20] and DBC [4], and transfer learning based approaches: PUL [6], PTGAN [31], SPGAN [3], CAMEL [37], MMFA [19], TJ-AIDL [30], HHL [45], ECN [46], MAR [38], PAUL [36], SSG [7], CR-GAN [1], CASCL [32], ”
Self-Similarity Grouping: A Simple Unsupervised Cross Domain Adaptation Approach for Person Re-Identification (2018, methodology): “Bagof-Words (BoW) [45] and local maximal occurrence (LOMO) [24], three unsupervised methods, including UMDL [29],PUL [10] and CAMEL [43], and six unsupervised domain adaptation methods, including PTGAN [40], SPGAN [8], TJ-AIDL [39], ARN [23], UDAP [33] and MAR [44].”
Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification (2020, methodology): “1 CLUSTERING-BASED UDA METHODS REVISIT State-of-the-art UDA methods (Fan et al., 2018; Lin et al., 2019; Zhang et al., 2019b; Yang et al., 2019) follow a similar general pipeline.”
Recognition From Web Data: A Progressive Filtering Approach (2018, methodology): “[23] improve the person re-identification [29] results by using a progressive method to train and rectify the model representation and clustering alternately.”
Improving Person Re-identification by Attribute and Identity Learning (2017, methodology): “ecently, some semi-supervised methods [19,37] and unsupervised ethods [22,38] has been proposed to address the data problem or re-ID.”
Dual-path Convolutional Image-Text Embeddings with Instance Loss (2017, unknown): “In this paper, in place of RNNs which are more commonly seen in image-text matching, we explore the usage of CNNs for text representation learning.”
Disentangled Person Image Generation (2017, methodology): “Euclidean distance is used to calculate the extracted features after l2-norm normalization [6].”
Self-paced Contrastive Learning with Hybrid Memory for Domain Adaptive Object Re-ID (2020, unknown): “Methods DukeMTMC-reID→Market-1501 Market-1501→DukeMTMC-reID mAP top-1 top-5 top-10 mAP top-1 top-5 top-10 PUL [8] TOMM’18 20.”
Group-aware Label Transfer for Domain Adaptive Person Re-identification (2021, background): “PUL [11] iteratively obtains the hard clustering labels and trains the model in a selftraining manner.”
Cluster Contrast for Unsupervised Person Re-Identification (2021, methodology): “The pseudo-label based unsupervised re-ID methods [14,15,17,30,47,4,49,57,65,62] use the supervised ImageNet pretrained backbone for clustering initialization.”