Re-identification Research Articles

Hybrid Contrastive Learning for Unsupervised Person Re-Identification

Unsupervised person re-identification (Re-ID) aims to learn discriminative features without human-annotated labels. Recently, contrastive learning provides a new prospect for unsupervised person Re-ID, and existing methods mainly constrain the feature similarity among easy sample pairs. However, the feature similarity among hard sample pairs is neglected, which causes suboptimal performance in unsupervised person Re-ID. In the paper, we propose a novel Hybrid Contrastive Model (HCM) to perform the identity-level contrastive learning and the image-level contrastive learning for unsupervised person Re-ID, which adequately explores the feature similarity among hard sample pairs. Specifically, for the identity-level contrastive learning, an identity-based memory is constructed to store pedestrian features. Accordingly, we define the dynamic contrast loss to identify identity information with dynamic factor for distinguishing hard/easy samples. As for the image-level contrastive learning, an image-based memory is established to store each image feature. We design the sample constraint loss to explore the similarity relationship between hard positive and negative sample pairs. Furthermore, we optimize the two contrastive learning processes in one unified framework to make use of their own advantages as so to constrain the feature distribution for extracting potential information. Extensive experiments prove that the proposed HCM distinctly outperforms the state-of-the-art methods.

Weakly Supervised Distribution Discrepancy Minimization Learning With State Information for Person Re-Identification

Weakly supervised person re-identification (Re-ID) is appealing to handle real-world tasks by using state information that is available without manual annotation. At present, most methods perform unsupervised cross domain (UCD) learning by transferring the knowledge from the labeled source domain to the unlabeled target domain, which results in poor performance due to the severe shift. To address this problem, in this paper, we utilize the tracklet and camera information as weak supervision to propose a distribution discrepancy minimization learning (DDML) model for UCD person Re-ID. In addition to aligning data distributions from the perspective of domain adaptation learning, two losses are developed from the view of neighborhood invariance exploration to optimize matching results. Specifically, to bridge the gap between domains, we propose a camera-distribution-based (CDB) loss to align pair-wise distance distributions. Furthermore, to alleviate the biased search within the target domain, we propose a ranking-confidence-based (RCB) loss to perform the mined neighborhood for intra-camera and inter-camera separately to explore a high degree of confidence neighbor relations. Extensive experiments on three challenging datasets demonstrate that applying our method to unlabeled target domain outperforms current weakly supervised methods for person Re-ID.

Occluded Person Re-Identification With Pose Estimation Correction and Feature Reconstruction

In the real-world surveillance system application, the accuracy of person re-identification (Re-ID) still suffers from occlusion. Occluded Re-ID aims to retrieve occluded person images from multiple cameras that do not overlap. To address this issue, we present a multi-branch feature enhancement network in this paper. It has the pose correction module (PCM), the feature reconstruction module (FRM), and the part align module (PAM). PCM builds the key-points confidence corrected mechanism of the self-adaptive weights learning before the part matching stage to strengthen part-level features in the non-occluded region and weaken ones in the occluded region. FRM reconstructs the feature distribution of inter-domain and class-domain using the maximum mean discrepancy to generalize the global features. The proposed feature separation mechanism enables the two branches to attend to distinct pedestrian feature. To increase the power of representation, PAM combines part-level and global features in an outer product. The experimental results on three holistic datasets, two partial datasets, and two occluded datasets demonstrate that our method is superior. It significantly outperforms state-of-the-art by 5.7% and 4.4% rank-1 scores on Occluded-Duke and Occluded-REID, respectively.

Attention Map Guided Transformer Pruning for Occluded Person Re-Identification on Edge Device

Due to its significant capability of modeling long-range dependencies, vision transformer (ViT) has achieved promising success in both holistic and occluded person re-identification (Re-ID) tasks. However, the inherent problems of transformers such as the huge computational cost and memory footprint are still two unsolved issues that will block the deployment of ViT based person Re-ID models on resource-limited edge devices. Our goal is to reduce both the inference complexity and model size without sacrificing the comparable accuracy on person Re-ID, especially for tasks with occlusion. To this end, we propose a novel attention map guided (AMG) transformer pruning method, which removes both redundant tokens and heads with the guidance of the attention map in a hardware-friendly way. We first calculate the entropy in the key dimension and sum it up for the whole map, and the corresponding head parameters of maps with high entropy will be removed for model size reduction. Then we combine the similarity and first-order gradients of key tokens along the query dimension for token importance estimation and remove redundant key and value tokens to further reduce the inference complexity. Comprehensive experiments on Occluded DukeMTMC and Market-1501 demonstrate the effectiveness of our proposals. For example, our proposed pruning strategy on ViT-Base enjoys <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">29.4%</b> <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">FLOPs</b> savings with <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0.2%</b> drop on Rank-1 and <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0.4%</b> improvement on mAP, respectively. Code and models have been made available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/NUST-Machine-Intelligence-Laboratory/AMG</uri> .

Win-Win by Competition: Auxiliary-Free Cloth-Changing Person Re-Identification

Recent person Re-IDentification (ReID) systems have been challenged by changes in personnel clothing, leading to the study of Cloth-Changing person ReID (CC-ReID). Commonly used techniques involve incorporating auxiliary information (e.g., body masks, gait, skeleton, and keypoints) to accurately identify the target pedestrian. However, the effectiveness of these methods heavily relies on the quality of auxiliary information and comes at the cost of additional computational resources, ultimately increasing system complexity. This paper focuses on achieving CC-ReID by effectively leveraging the information concealed within the image. To this end, we introduce an Auxiliary-free Competitive IDentification (ACID) model. It achieves a win-win situation by enriching the identity (ID)-preserving information conveyed by the appearance and structure features while maintaining holistic efficiency. In detail, we build a hierarchical competitive strategy that progressively accumulates meticulous ID cues with discriminating feature extraction at the global, channel, and pixel levels during model inference. After mining the hierarchical discriminative clues for appearance and structure features, these enhanced ID-relevant features are crosswise integrated to reconstruct images for reducing intra-class variations. Finally, by combing with self- and cross-ID penalties, the ACID is trained under a generative adversarial learning framework to effectively minimize the distribution discrepancy between the generated data and real-world data. Experimental results on four public cloth-changing datasets (i.e., PRCC-ReID, VC-Cloth, LTCC-ReID, and Celeb-ReID) demonstrate the proposed ACID can achieve superior performance over state-of-the-art methods. The code is available soon at: https://github.com/BoomShakaY/Win-CCReID.

A Real-Time Memory Updating Strategy for Unsupervised Person Re-Identification

Recently, clustering-based methods have been the dominant solution for unsupervised person re-identification (ReID). Memory-based contrastive learning is widely used for its effectiveness in unsupervised representation learning. However, we find that the inaccurate cluster proxies and the momentum updating strategy do harm to the contrastive learning system. In this paper, we propose a real-time memory updating strategy (RTMem) to update the cluster centroid with a randomly sampled instance feature in the current mini-batch without momentum. Compared to the method that calculates the mean feature vectors as the cluster centroid and updating it with momentum, RTMem enables the features to be up-to-date for each cluster. Based on RTMem, we propose two contrastive losses, i.e., sample-to-instance and sample-to-cluster, to align the relationships between samples to each cluster and to all outliers not belonging to any other clusters. On the one hand, sample-to-instance loss explores the sample relationships of the whole dataset to enhance the capability of density-based clustering algorithm, which relies on similarity measurement for the instance-level images. On the other hand, with pseudo-labels generated by the density-based clustering algorithm, sample-to-cluster loss enforces the sample to be close to its cluster proxy while being far from other proxies. With the simple RTMem contrastive learning strategy, the performance of the corresponding baseline is improved by 9.3% on Market-1501 dataset. Our method consistently outperforms state-of-the-art unsupervised learning person ReID methods on three benchmark datasets. Code is made available at:https://github.com/PRIS-CV/RTMem.

Self-Regulation Feature Network for Person Reidentification

The performance of person reidentification (ReID) has been significantly improved with the development of convolutional neural networks in the past few years. Most existing ReID methods focus on modifying the backbone network architecture to learn feature representations of persons; however, there are only a few methods that explore the potential performance of the network architecture to relearn features. In this article, we propose a self-regulation feature (SRF) scheme based on the network model, called SRFnet, to address the problem of the limited ability of the model for learning person features. The proposed SRFnet utilizes global branches to supervise local branches and deeply mine potential features of the network from the perspective of optimization to obtain more distinguished feature representations of persons. Different from methods that add an attention mechanism, our SRFnet can self-adjust feature learning by only adding loss functions between the local and global features of the network, and it does not require any additional convolution parameters. Extensive experiments on three public datasets show that the proposed SRFnet method achieves the results of mean average precision (mAP)/Rank-1 as 93.65%/95.28% on Market1501, 89.65%/91.79% on DuckMTMC-reID, 83.79%/82.00% on CUHK03 (labeled), and 81.92%/80.21% on CUHK03 (detected) datasets, respectively. Furthermore, experimental results on cross-domain datasets also demonstrate the effectiveness of the proposed SRFnet.

A Novel Mix-Normalization Method for Generalizable Multi-Source Person Re-Identification

Person re-identification (Re-ID) has achieved great success in the supervised scenario. However, it is difficult to directly transfer the supervised model to arbitrary unseen domains due to the model overfitting to the seen source domains. In this paper, we aim to tackle the generalizable multi-source person Re-ID task (i.e., there are multiple available source domains, and the testing domain is unseen during training) from the data augmentation perspective, thus we put forward a novel method, termed MixNorm. It consists of domain-aware mix-normalization (DMN) and domain-aware center regularization (DCR). Different from the conventional data augmentation, the proposed domain-aware mix-normalization enhances the diversity of features during training from the normalization perspective of the neural network, which can effectively alleviate the model overfitting to the source domains, so as to boost the generalization capability of the model in the unseen domain. To further promote the efficacy of the proposed DMN, we exploit the domain-aware center regularization to better map the diversely generated features into the same space. Extensive experiments on multiple benchmark datasets validate the effectiveness of the proposed method and show that the proposed method can outperform the state-of-the-art methods. Besides, further analysis also reveals the superiority of the proposed method.

Exploring High-order Spatio-temporal Correlations from Skeleton for Person Re-identification.

Person re-identification (Re-ID) has become a hot research topic due to its widespread applications. Conducting person Re-ID in video sequences is a practical requirement, in which the crucial challenge is how to pursue a robust video representation based on spatial and temporal features. However, most of the previous methods only consider how to integrate part-level features in the spatio-temporal range, while how to model and generate the part-correlations is little exploited. In this paper, we propose a skeleton-based dynamic hypergraph framework, namely Skeletal Temporal Dynamic Hypergraph Neural Network (ST-DHGNN) for person Re-ID, which resorts to modeling the high-order correlations among various body parts based on a time series of skeletal information. Specifically, multi-shape and multi-scale patches are heuristically cropped from feature maps, constituting spatial representations in different frames. A joint-centered hypergraph and a bone-centered hypergraph are constructed in parallel from multiple body parts (i.e., head, trunk, and legs) with spatio-temporal multi-granularity in the entire video sequence, in which the graph vertices representing regional features and hyperedges denoting relationships. Dynamic hypergraph propagation containing the re-planning module and the hyperedge elimination module is proposed to better integrate features among vertices. Feature aggregation and attention mechanisms are also adopted to obtain a better video representation for person Re-ID. Experiments show that the proposed method performs significantly better than the state-of-the-art on three video-based person Re-ID datasets, including iLIDS-VID, PRID-2011, and MARS.

An Image–Text Dual-Channel Union Network for Person Re-Identification

Multiple people can have similar appearances and portions of images can be occluded or have viewpoint changes in real scenarios, causing the increased difficulty of person re-identification (Re-ID). To address these problems, we propose a dual-channel person Re-ID algorithm that integrates person Re-ID image-text pairs into the classification network for end-to-end learning. We construct an image channel and a text channel, and subsequently extract visual information and text information using a convolutional neural network and a simple recurrent units network, respectively. The text information is used to assist in the learning of visual information, consequently improving the robustness of the visual information. In addition, the visual features are divided into two branches to calculate the global and local features. Global features focus on the overall appearance of a person, whereas local features provide more fine-grained detail. Text information is more accurate and reliable, and it is thus more robust to occlusion and viewpoint changes. Visual information complemented by text information can describe a person more accurately and reliably. Extensive experiments demonstrate our method achieves state-of-the-art performance.

Review of Visible-Infrared Cross-Modality Person Re-Identification

Person re-identification (ReID) is a sub-problem under image retrieval. It is a technology that uses computer vision to identify a specific pedestrian in a collection of pictures or videos. The pedestrian image under cross-device is taken from a monitored pedestrian image. At present, most ReID methods deal with the matching between visible and visible images, but with the continuous improvement of security monitoring system, more and more infrared cameras are used to monitor at night or in dim light. Due to the image differences between infrared camera and RGB camera, there is a huge visual difference between cross-modality images, so the traditional ReID method is difficult to apply in this scene. In view of this situation, studying the pedestrian matching between visible and infrared modalities is particularly crucial. Visible-infrared person re-identification (VI-ReID) was first proposed in 2017, and then attracted more and more attention, and many advanced methods emerged.

Dual Level Adaptive Weighting for Cloth-Changing Person Re-Identification.

For the long-term person re-identification (ReID) task, pedestrians are likely to change clothes, which poses a key challenge in overcoming drastic appearance variations caused by these cloth changes. However, analyzing how cloth changes influence identity-invariant representation learning is difficult. In this context, varying cloth-changed samples are not adaptively utilized, and their effects on the resulting features are overshadowed. To address these limitations, this paper aims to estimate the effect of cloth-changing patterns at both the image and feature levels, presenting a Dual-Level Adaptive Weighting (DLAW) solution. Specifically, at the image level, we propose an adaptive mining strategy to locate the cloth-changed regions for each identity. This strategy highlights the informative areas that have undergone changes, enhancing robustness against cloth variations. At the feature level, we estimate the degree of cloth-changing by modeling the correlation of part-level features and re-weighting identity-invariant feature components. This further eliminates the effects of cloth variations at the semantic body part level. Extensive experiments demonstrate that our method achieves promising performance on several cloth-changing datasets. Code and models are available at https: //github.com/fountaindream/DLAW.

A Dual-Branch Network for Few-Shot Vehicle Re-Identification With Enhanced Global and Local Features

Traditional vehicle Re-Identification (Re-ID) methods mainly rely on large-size training samples to achieve better results. However, obtaining abundant training samples is challenging for applications of these methods to few-shot vehicle Re-ID in real-world traffic scenes. To address this challenge, we propose a dual-branch network (DB-Net) for few-shot vehicle Re-ID with enhanced global and local features. Our proposed method reduces the dependence of vehicle Re-ID network on a large number of training samples by improving its feature expression ability and feature quality, rather than synthesizing vehicle images in traditional methods. The proposed DB-Net, composed of global and local branches, extracts global appearance features and local detail features of the vehicle. A global feature optimization module in the DB-Net mines and maintains more global semantic information using convolution and concat operations. And a feature screening module selects the most relevant local features based on Mutual Information and Shell Sorting to enhance local features. Additionally, a local attention module assigns adaptive weights to enhance salient local regions. Our approach is evaluated on a new dataset (Veri-FS) with small sample sizes and poor illumination conditions. It outperforms state-of-the-art methods on VehicleID, VeRi-776, and Veri-FS datasets, demonstrating its effectiveness in few-shot vehicle Re-ID.

Complementary Attention-Driven Contrastive Learning With Hard-Sample Exploring for Unsupervised Domain Adaptive Person Re-ID

Unsupervised domain adaptive (UDA) methods for person re-identification (Re-ID) aim to transfer the knowledge of the labeled source domain to the unlabeled target domain without further annotations, which is challenging due to the drift of label distribution and the missing of target domain labels. Improving the clustering accuracy of pseudo-labels can help the model fit the target domain. However, the errors of pseudo-label noise will be accumulated during training, which is harmful to the model performance. Moreover, the hard samples can lead to a large gap between intra-class features and a small gap between inter-class features. To address these problems, this paper proposes a complementary attention-driven contrastive learning with hard-sample exploring (CACHE) algorithm. In CACHE, on one hand, the complementary attention module is used to improve the discriminability of the features. The obtained discriminative features can reduce noisy pseudo-labels and improve the clustering accuracy of pseudo labels; On the other hand, we explore the hard samples based on the instance relationship and cluster relationship for contrastive learning. This way can make the cluster more compact. Extensive experiments on three large-scale person re-identification benchmarks demonstrate the effectiveness of the proposed method, which significantly outperforms state-of-the-art methods in terms of mAP and CMC.

RelationTrack: Relation-Aware Multiple Object Tracking With Decoupled Representation

Existing online multiple object tracking (MOT) algorithms often consist of two subtasks, detection and re-identification (ReID). In order to enhance the inference speed and reduce the complexity, current methods commonly integrate these double subtasks into a unified framework. Nevertheless, detection and ReID demand diverse features. This issue results in an optimization contradiction during the training procedure. With the target of alleviating this contradiction, we devise a module named Global Context Disentangling (GCD) that decouples the learned representation into detection-specific and ReID-specific embeddings. As such, this module provides an implicit manner to balance the different requirements of these two subtasks. Moreover, we observe that preceding MOT methods typically leverage local information to associate the detected targets and neglect to consider the global semantic relation. To resolve this limitation, we develop a module, referred to as Guided Transformer Encoder (GTE), by combining the powerful reasoning ability of Transformer encoder and deformable attention. Unlike previous works, GTE avoids analyzing all the pixels and only attends to capture the relation between query nodes and a few self-adaptively selected key samples. Therefore, it is computationally efficient. Extensive experiments have been conducted on the MOT16, MOT17 and MOT20 benchmarks to demonstrate the superiority of the proposed MOT framework, namely RelationTrack. The experimental results indicate that RelationTrack has surpassed preceding methods significantly and established a new state-of-the-art performance, e.g., IDF1 of 70.5% and MOTA of 67.2% on MOT20.

Consistent Discrepancy Learning for Intra-Camera Supervised Person Re-Identification

Since annotating pedestrians across different views is extremely costly, intra-camera supervised person re-identification (ReID) aims to learn a ReID model from the intra-view labeled data. Under this setting, the most challenge lies in learning a view-invariant feature embedding in the absence of the cross-view annotations. Previous works focus on assigning a pseudo identity label for each image based on the feature similarity and learn view-invariant features by classification loss. However, because of the cross-view variations in lighting, background, etc., the pseudo labels are often noisy, and therefore not reliable for classification. In this paper, we explore learning a consistent discrepancy for pairwise images. Our main idea is that the discrepancy between pedestrian images should be consistent across different views regardless of view change so that it mainly depicts the identity difference. Due to the lack of cross-view annotations, we project images into different views and obtain likelihood prototypes for cross-view learning. These likelihood prototypes are used to measure the discrepancies between pairwise images under different views. And then, we propose an intra-view discrepancy preservation module to enforce the discrepancy to be view-consistent so as to encourage the model to distinguish the images based on the identities regardless of view change. Extensive experiments on multiple datasets show that our method outperforms existing related methods by clear margins and our method is comparable to supervised counterparts. Code will be made publicly available.

CNN with Embedding Transformers for Person Reidentification

For person reidentification (ReID), most slicing methods (such as part‐based convolutional baseline (PCB) and AlignedReID) introduce a lot of background devoid of pedestrian parts, resulting in the cross‐aliasing of features in the deep network. Besides, the resulting component features are not perfectly aligned with each other, thus affecting model performance. We propose a convolutional neural network (CNN) with embedding transformers (CET) person ReID network architecture based on the respective advantages of CNN and transformer. In CET, first, the residual transformer (RT) structure is first embedded in the backbone network of CNN to obtain a feature extractor, named transformers in CNN. The feature aliasing phenomenon is improved by utilizing transformer’s advantage in grasping the relevance of global information. Second, a feature fuse with learnable vector structure for fusing the output vector is added to the output of the transformer at the end of the network. A two branches loss structure is designed to balance the two different fusion strategies. Finally, the self‐attention mechanism in transformer is used for automatic part alignment of human body parts to solve the part alignment problem caused by inaccurate detection frames. The experimental results show that CET network architecture achieves better performance than PCB and some other block‐slicing methods.

Unsupervised Domain Adaptation for Person Re-Identification Via Individual-Preserving and Environmental-Switching Cyclic Generation

Unsupervised domain adaptation for person re-identification (Re-ID) suffers severe domain discrepancies between source and target domains. To reduce the domain shift caused by the changes of context, camera style, or viewpoint, existing methods in this field fine-tune and adapt the Re-ID model with augmented samples, either translating source samples to the target style or assigning pseudo labels to the target. The former methods may lose identity details but keep redundant source background during translation. In contrast, the latter techniques may give noisy labels when the model meets the unseen background and person pose. We mitigate the domain shift in the former translation direction by cyclically decoupling environment and identity-related features. We propose a novel individual-preserving and environmental-switching cyclic generation network (IPES-GAN). Our network has the following distinct features: 1) <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Decoupled features instead of fused features:</i> we encode the images into an individual part and an environmental part, which are proved beneficial to generation and adaptation; 2) <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Cyclic generation instead of one-step adaptive generation</i> . We swap source and target environment features to generate cross-domain images with preserved identity-related features conditioned with source (target) background features and then changed again to generate back the input image so that cyclic generation runs in a self-supervised way. Experiments carried out on two significant benchmarks: Market-1501 and DukeMTMC-Reid, reveal state-of-the-art performance.

Multi-granularity graph pooling for video-based person re-identification

The video-based person re-identification (ReID) aims to identify the given pedestrian video sequence across multiple non-overlapping cameras. To aggregate the temporal and spatial features of the video samples, the graph neural networks (GNNs) are introduced. However, existing graph-based models, like STGCN, perform the mean/max pooling on node features to obtain the graph representation, which neglect the graph topology and node importance. In this paper, we propose the graph pooling network (GPNet) to learn the multi-granularity graph representation for the video retrieval, where the graph pooling layer is implemented to downsample the graph. We construct a multi-granular graph by using node features learned from backbone, then implement multiple graph convolutional layers to perform the spatial and temporal aggregation on nodes. To downsample the graph, we propose a multi-head full attention graph pooling (MHFAPool) layer, which integrates the advantages of existing node clustering and node selection pooling methods. Specifically, MHFAPool first learns a full attention matrix for each pooled node, then obtains the principal eigenvector of the attention matrix via the power iteration algorithm, finally takes the softmax of the principal eigenvector as the aggregation coefficients. Extensive experiments demonstrate that our GPNet achieves the competitive results on four widely-used datasets, i.e., MARS, DukeMTMC-VideoReID, iLIDS-VID and PRID-2011.

Two-phase self-supervised pretraining for object re-identification

Self-supervised pretraining (e.g., MAE) based on reconstructing masked image patches has become increasingly popular in recent months. It commonly pretrains models on the large-scale ImageNet dataset with a considerably wider range of image categories, and then finetunes on downstream tasks such as classification, detection, segmentation, etc. However, we discover that for finetuning domain-specific tasks usually with a few or limited semantic spaces (e.g., object re-identification (ReID)), the model pretrained on a wider variety of sources (e.g., ImageNet) fails to be a good initialization. To address the problem, we propose a second-phase self-supervised pretraining on the specific domain, based on the self-supervised ImageNet pretraining models in the first phase. We perform studies on 2 domains of object ReID, i.e., person and vehicle ReID, where person ReID domain involves four datasets (Market-1501, DukeMTMC, MSMT17, Occluded-Duke) and vehicle ReID domain contains 2 datasets (VeRi-776, VehicleID). Through extensive experiments, we observe significant improvements in every single dataset based on the second-phase self-supervised pretraining over the entire unlabeled domain data. Specifically in person ReID task, compared to the self-supervised model in the first phase, our approach achieves +8.5%/+11.5%/+13.3%/+17.2% mAP improvements in Market-1501, DukeMTMC, MSMT17 and Occluded-Duke datasets, respectively. Furthermore, it also outperforms the supervised pretraining baselines, achieving improved performance by +1.5%/+4.3%/+3.9%/+4.2% mAP. We observe similar inspiring improvements in vehicle ReID tasks. Additionally, we also perform the second-phase pretraining on the specific tasks using its own unlabeled data, namely task-specific pretraining (TSP), which also consistently shows encouraging results across object ReID tasks.