Multi-view Based Pose Alignment Method for Person Re-identification

Many features have been proposed to improve the accuracy of person re-identification. Due to the illumination and viewpoint changes between different cameras, individual feature is less discriminative to separate different persons. In this paper, we propose a saliency-weighted feature descriptor and global-local fusion optimization for person re-identification. Firstly, the weights on pixels are calculated via saliency detection method, then the computed weights are integrated into local maximal occurrence (LOMO) feature descriptor. Secondly, the saliency weights are used to update the metric learning distance in training so that we can learn a new metric matrix for testing. And then, the whole person image is divided into upper and lower halves. A novel global-local fusion method is proposed to combine local and global regions together in the most appropriate way. After that an optimization algorithm is proposed to learn the weights among upper half, lower half and the whole image. According to those weights, a final fused distance is obtained. Experimental results show that the proposed method outperforms many state-of-the-art person re-identification methods.

This work introduces a novel method for person re-identification using embedded smart cameras. State-of-the-art methods address the re-identification problem using global and local features, metric learning and feature transformation algorithms. Such methods require advanced systems with high computational capabilities. Nowadays, there is a growing interest in security applications using embedded cameras. Motivated by this we propose to study a new system that addresses the challenges posed by the reidentification problem using devices (e.g. smartphones, etc.) that have limited resources. In this work we introduce a novel client-server system that exploits a feature learning method to achieve a two-fold objective: (i) maximize the re-identification performance over time and (ii) reduce the required computational costs. In the training phase, state-of-the-art features are selected considering both the device capabilities and re-identification performance. During the detection phase, the re-identification performance are maximized by selecting the best features for a given input image. To demonstrate the performance of the proposed method we conduct the experiments using different mobile devices. Statistics about feature extraction and feature matching are presented together with re-identification results.

To fully pay attention to identity-sensitive feature information and utilize the correlations of inter-attributes and attributes-body parts, this paper proposes a person re-identification (re-ID) method based on the construction of graph convolutional network (GCN) with crucial attribute feature and body parts. First, it establishes the multiscale context-aware network (MSCAN) using dilated convolution with different expansion ratios, which can learn multiscale context information and obtain diversified global features. Subsequently, the human parsing model is utilized to extract the body part features. According to the attribute importance degree, the paper constructs low-dimensional GCN integrating the vital attributes and body parts of person descriptions to obtain discriminative local features. Finally, based on attribute prediction, it reduces the range of the images to be matched with discriminating possible objects from query images, thereby simplifying retrieval process. The experimental results demonstrate that the novel designed method can effectively improve person re-ID performance and achieve competitive evaluation results on typical public testing datasets.

To address challenges related to the inadequate representation and inaccurate discrimination of pedestrian attributes, we propose a novel method for person re-identification, which leverages global feature learning and classification optimization. Specifically, this approach integrates a Normalization-based Channel Attention Module into the fundamental ResNet50 backbone, utilizing a scaling factor to prioritize and enhance key pedestrian feature information. Furthermore, dynamic activation functions are employed to adaptively modulate the parameters of ReLU based on the input convolutional feature maps, thereby bolstering the nonlinear expression capabilities of the network model. By incorporating Arcface loss into the cross-entropy loss, the supervised model is trained to learn pedestrian features that exhibit significant inter-class variance while maintaining tight intra-class coherence. The evaluation of the enhanced model on two popular datasets, Market1501 and DukeMTMC-ReID, reveals improvements in Rank-1 accuracy by 1.28% and 1.4%, respectively, along with corresponding gains in the mean average precision (mAP) of 1.93% and 1.84%. These findings indicate that the proposed model is capable of extracting more robust pedestrian features, enhancing feature discriminability, and ultimately achieving superior recognition accuracy.

Unsupervised Domain Adaptation (UDA) methods for person Re-Identification (Re-ID) rely on target domain samples to model the marginal distribution of the data. To deal with the lack of target domain labels, UDA methods leverage information from labeled source samples and unlabeled target samples. A promising approach relies on the use of unsupervised learning as part of the pipeline, such as clustering methods. The quality of the clusters clearly plays a major role in methods performance, but this point has been overlooked. In this work, we propose a multi-step pseudo-label refinement method to select the best possible clusters and keep improving them so that these clusters become closer to the class divisions without knowledge of the class labels. Our refinement method includes a cluster selection strategy and a camera-based normalization method which reduces the within-domain variations caused by the use of multiple cameras in person Re-ID. This allows our method to reach state-of-the-art UDA results on DukeMTMC-Market1501 (source-target). We surpass state-of-the-art for UDA Re-ID by 3.4% on Market1501-DukeMTMC datasets, which is a more challenging adaptation setup because the target domain (DukeMTMC) has eight distinct cameras. Furthermore, the camera-based normalization method causes a significant reduction in the number of iterations required for training convergence.

For person re-identification the appearance of photographs taken by different cameras at different viewpoints varies considerably. In this paper, a person re-identification method based on multi-view and attention mechanism is proposed. Firstly, IBN-Net is introduced as the backbone network based on ResNet50, and CA attention mechanism is embedded into the backbone network to obtain person information with more discriminative features. Then, global, spatial and channel features are combined in order to obtain multi-scale feature information and local features of people. It is also considered that the person is a non-rigid structure, and its pose and viewpoint changes can cause huge differences in person appearance features from one camera to another and can ultimately affect the accuracy of recognition. Therefore, a mesh-based view generator is introduced to generate multi-view person images, and multi-view image enhancement methods are utilized to reduce the sensitivity of the network to viewpoint. Finally, the model is optimized by introducing three loss functions. Experiments conducted on three publicly available benchmark datasets Market-1501, DukeMTMC-ReID and CUHK03, the experimental results show that our proposed model is at a high level in both single-domain and cross-domain tasks.

To improve the accuracy of person re-identification and reduce the size of models, a lightweight and accurate (LWA) unsupervised method is proposed. Firstly, a lightweight image classifier (LWC) is proposed. It combines DenseNet's convolutional layer and batch normalization layer to simplify its model structure. Then the spatio-temporal information of datasets is added to form a powerful classifier. Secondly, an optimization learning algorithm is proposed. A feeble classifier (LWB) is implemented by constructing a binary network. For the training of the network, the ranking information of a powerful classifier is used to enhance the recognition capability of the LWB. Furthermore, spatio-temporal information of datasets is added again to construct LWA classifier. The results show that the model size of LWB is about 80% smaller than that of the Siamese network model on the CUHK01 and VIPeR datasets. At the same time, the recognition accuracy of the LWA classifier is slightly improved on the Market-1501 and DukeMTMC target datasets.

Prevalent nighttime person re-identification (ReID) methods typically combine image relighting and ReID networks in a sequential manner. However, their performance (recognition accuracy) is limited by the quality of relighting images and insufficient collaboration between image relighting and ReID tasks. To handle these problems, we propose a novel Collaborative Enhancement Network called CENet, which performs the multilevel feature interactions in a parallel framework, for nighttime person ReID. In particular, the designed parallel structure of CENet can not only avoid the impact of the quality of relighting images on ReID performance, but also allow us to mine the collaborative relations between image relighting and person ReID tasks. To this end, we integrate the multilevel feature interactions in CENet, where we first share the Transformer encoder to build the low-level feature interaction, and then perform the feature distillation that transfers the high-level features from image relighting to ReID, thereby alleviating the severe image degradation issue caused by the nighttime scenario while avoiding the impact of relighting images. In addition, the sizes of existing real-world nighttime person ReID datasets are limited, and large-scale synthetic ones exhibit substantial domain gaps with real-world data. To leverage both small-scale real-world and large-scale synthetic training data, we develop a multi-domain learning algorithm, which alternately utilizes both kinds of data to reduce the inter-domain difference in training procedure. Extensive experiments on two real nighttime datasets, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Night600</i> and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">RGBNT201_rgb</i> , and a synthetic nighttime ReID dataset are conducted to validate the effectiveness of CENet. We release the code and synthetic dataset at: https://github.com/Alexadlu/CENet.

Most existing person re-identification (re-id) methods are designed based on the artificial closed-set assumption that the probe and gallery identities are exactly overlapped with a small search pool. This leads to poor scalability in real-world applications where the task is often to re-id a small set of target people (i.e., watch-list) among a large search pool with unknown ID overlap, namely, an open-set deployment setting. In this paper, we firstly propose a new person re-id setting called Watch-List based Open-Set (WLOS) person re-id, which is characterised by the above open-set deployment and a watch-list available at the training stage. Then, we address such a under-studied WLOS problem by formulating a novel Task Dedicated Deep Hashing (TDDH) approach which learning a purpose-specific deep hash model particularly for the given target people in an efficient end-to-end manner. Extensive experiments on three large-scale re-id benchmarks are conducted to demonstrate the advantages and superiority of the TDDH over a wide range of the state-of-the-art hashing and re-id methods under the more realistic open-set setting.

This paper proposes a method for video-based person re-identification. Motivated by the capacity of Recurrent Feature Aggregation Network (RFA-Net) that allows to aggregate the image-level features over time-steps and to yield a sequence level feature. Our work improves the person re- identification performance by (1) proposing to use Gaussian of Gaussian (GOG) - a more powerful feature at image level instead of LBP-Color as in the original work and (2) applying metric learning techniques for person matching. To analyze the role of image feature and metric learning, we have performed extensive experiments on two public benchmark datasets including PRID 2011 and iLIDS- VID. The proposed solution obtains impressive results on the PRID 2011 dataset. The matching rate at rank 1 is increased by 16.8% compared to that of original method. For the iLID-SVID dataset, the matching rate at rank 1 of proposed method is increased 1.1% whereas these values at the latter ranks are higher than those of the original method by approximately 4%. In additional, the mean and standard deviation of the rank where the correct matching is found of the proposed method on the PRID 2011 and the iLIDS-VID are 2.32, 4.33 and 4.61, 7.67 while those of the original work are 3.50, 5.42 and 8.03, 17.52, respectively. These obtained results emphasize the robustness of the proposed method compared to the original one.

Joint graph regularized dictionary learning and sparse ranking for multi-modal multi-shot person re-identification

In recent years, considerable progress has been witnessed in the person re-identification (Re-ID). However, in a more realistic long-term scenario, the appearance shift arising from the clothes-changing inevitably deteriorates the conventional methods that heavily depend on the clothing color. Although the current clothes-changing person Re-ID methods introduce external human knowledge (i.e, contour, mask) and sophisticated feature decoupling strategy to alleviate the clothing shift, they still face the risk of overfitting to clothing due to the limited clothing diversity of training set. To more efficiently and effectively promote the clothes-irrelevant feature learning, we present a novel joint Identity-aware Mixstyle and Graph-enhanced Prototype method for clothes-changing person Re-ID. Specifically, by treating the cloth-changing as fine-grained domain/style shift, the identity-aware mixstyle (IMS) is proposed from the perspective of domain generalization, which mixes the instance-level feature statistics of samples within each identity to synthesize novel and diverse clothing styles, while retaining the correspondence between synthesized samples and latent label space. By incorporating the IMS module, the more diverse styles can be exploited to train a clothing-shift robust model. To further reduce the feature discrepancy caused by clothing variations, the graph-enhanced prototype constraint (GEP) module is proposed to explore the graph similarity structure of style-augmented samples across memory bank to build informative and robust prototypes, which serve as powerful exemplars for better clothing-irrelevant metric learning. The two modules are integrated into a joint learning framework and benefit each other. The extensive experiments conducted on clothes-changing person Re-ID datasets validate the superiority and effectiveness of our method. In addition, our method also shows good universality and corruption robustness on other Re-ID tasks.

The general Re-ID works including datasets and networks have an assumption that the pedestrians do not change their appearances throughout the research. The lack of diversity about the appearance of the Re-ID datasets significantly limits the performance of models in cross-appearance Re-ID. Therefore, this paper conduct data augmentation in the appearance dimension to the datasets by the generation framework CAG(cross-appearance images generation) to support the studies about cross-appearance Re-ID. To demonstrate the effectiveness of the generation framework, we selected several classical and SOTA Re-ID models and conducted experiments on three commonly used cross-appearance Re-ID datasets, NKUP+IPRCC/DeepChange. The results show that the cross-appearance Re-ID images generated by CAG can help the models to obtain the robust pedestrian feature, and the diversity of appearance is a universal method for cross-appearance person re-identification.

Researches on person re-identification have become more prevalent as a core technique in visual surveillance systems. While a large volume of person re-identification (Re-ID) methods adopt experimental settings with a large patch size, relatively small ones are not much considered. Thus, this work focuses on investigating comparison of person Re-ID performance when color and shape features-based are adopted for small patch size images with low resolution. First, the deep decompositional network is used to divide the person into upper and lower body parts. Then, color and shape features are extracted. Finally, using single or combination of features, similarity-based ranked matching scores are computed. The person Re-ID performance is evaluated based on VIPeR dataset. From the experiment results, we found that color-based feature is better than shape-based features, and the combination of color and shape features-based can be meaningful.

Existing person re-identification (re-id) methods mostly assume the availability of large-scale identity labels for model learning in any target domain deployment. This greatly limits their scalability in practice. To tackle this limitation, we propose a novel Instance-Guided Context Rendering scheme, which transfers the source person identities into diverse target domain contexts to enable supervised re-id model learning in the unlabelled target domain. Unlike previous image synthesis methods that transform the source person images into limited fixed target styles, our approach produces more visually plausible, and diverse synthetic training data. Specifically, we formulate a dual conditional generative adversarial network that augments each source person image with rich contextual variations. To explicitly achieve diverse rendering effects, we leverage abundant unlabelled target instances as contextual guidance for image generation. Extensive experiments on Market-1501, DukeMTMC-reID and CUHK03 benchmarks show that the re-id performance can be significantly improved when using our synthetic data in cross-domain re-id model learning.