Person Matching Research Articles

Multi deep invariant feature learning for cross-resolution person re-identification

Person re-identification (Re-ID) is focused on identifying and matching the same pedestrian when captured across various surveillance cameras. However, variations in camera performance and the distances between pedestrians and cameras may result in capturing images of the same person at different resolutions. This specific issue, referred to as cross-resolution person re-identification (CRReID), presents considerable difficulties in achieving accurate person Re-ID. To address this issue, we propose a Multi-level and scale Deep Invariant Feature learning Framework (MDIFF), which effectively tackles the problem of cross-resolution person matching. Our MDIFF reconstructs person features at the shallow layers to alleviate information gaps and extracts resolution-invariant features at the deep layers for cross-resolution person matching. First, to mitigate information loss in resolution-invariant features, we propose a Dual Input Feature Reconstruction (DIFR) structure incorporating dual-stream input and a lightweight decoder, constrained by degradation loss and image reconstruction loss. Second, we propose a Multi-level Global–Local feature Interaction and Fusion (MGLIF) module to enhance the invariant features of persons and obtain deep invariant representations, making the final representation more robust to resolution changes and more discriminative. Finally, to make the feature distribution of the same identity more compact across different resolutions, we propose a cross-resolution joint loss optimization strategy, including cross-resolution triplet loss, cross-resolution center loss, and identity loss. Our comprehensive experimental results demonstrate the superior performance and efficacy of our MDIFF, outperforming current state-of-the-art methods across various CRReID benchmark datasets. Our code is available at https://github.com/MiSanl/MDIFF-for-CRReID.

An empirical study of the personality factors and individual characteristics of older adult participation in third age education

ABSTRACT In China, the innovative development of education for older people is an important response to the aging population, and an important way to meet the spiritual, cultural, and social needs of older people. Current academic research mainly focuses on the social factors that affect the participation of older people in education and there are few empirical studies on the individual psychological characteristics of older adult learning behavior. This study takes a positive psychology approach to the exploration of the factors influencing the participation of older adults in education in a number of universities of the Third Age in Fujian Province, China. The study is based on curriculum preferences and frequency of participation in extracurricular activities. The results show that the personality traits of extraversion and neuroticism, as well as gender, age, level of educational attainment, and family economic status significantly affect frequency of participation in extracurricular activities and the curriculum preferences of older learners. The results of this study suggest that we can strengthen personality matching to promote the learning enthusiasm of older people in education, and adhere to the demand-oriented to promote the Third age education curriculum personalized. Furthermore, we should increase the scope for popularizing Third age education, improve the accessibility of education for the older people, accelerate the integration of older adult education with the Internet, innovate the form of older adult education, and then promote the development of customized curricula.

Context-aided unicity matching for person re-identification

Most existing person re-identification methods compute the matching relations between person images based on the similarity ranking. It lacks the global viewpoint and context consideration, inevitably leading to ambiguous matching results and sub-optimal performance. Based on a natural assumption that images belonging to the same identity should not match with images belonging to different identities across views, called the unicity of person matching on the identity level, we propose an end-to-end person unicity matching architecture to learn and refine the person matching relations. We first employ the image samples’ context in feature space to generate the initial soft matchings using graph neural networks, and then utilize the samples’ global context to refine the soft matchings and reach the matching unicity by bipartite graph matching. Considering real-world applications, we further develop a fast algorithm without losing performance. Experimental results on five public benchmarks show the superiority of the proposed method.

Which scientist are you? Creating self-outgroup overlap with a scientist through a personality matching game

Based in intergroup contact theory, we investigated how messaging about shared characteristics affects perceived closeness with scientists (i.e., self-outgroup overlap). In an online study, participants ($N=486$) played a personality matching game that matched them with a real scientist, then they responded to a survey. We replicated the study at a zoo ($N=63$) to examine implementation as a facilitated game. Self-scientist overlap improved in the online setting; in the in situ setting, trust increased, but not self-scientist overlap. Findings suggest that learning about how one scientist is similar to one's self can increase perceived closeness to scientists overall.

MHSA-Net: Multihead Self-Attention Network for Occluded Person Re-Identification.

This article presents a novel person reidentification model, named multihead self-attention network (MHSA-Net), to prune unimportant information and capture key local information from person images. MHSA-Net contains two main novel components: multihead self-attention branch (MHSAB) and attention competition mechanism (ACM). The MHSAB adaptively captures key local person information and then produces effective diversity embeddings of an image for the person matching. The ACM further helps filter out attention noise and nonkey information. Through extensive ablation studies, we verified that the MHSAB and ACM both contribute to the performance improvement of the MHSA-Net. Our MHSA-Net achieves competitive performance in the standard and occluded person Re-ID tasks.

Efficient Person Reidentification for IoT-Assisted Cyber–Physical Systems

The main objective of this study is to propose a cyber-physical system (CPS) based person re-identification framework for smart surveillance. The Internet of Things (IoT) based interconnected vision sensors in smart cities are considered essential elements of a CPS, and contribute significantly to urban security. However, the re-identification of targeted persons using emerging edge AI techniques still faces certain challenges. To improve efficiency at the edge and overcome the traditional sensing of video cameras, we employed an AI-based person re-identification framework for CPS that is functional in IoT environments. In addition, we present dual attention dilated network (DADNet), which integrates an energy-efficient convolutional neural network (CNN) with a self-attention module to substantially improve the person matching probability. Further, we applied dual feature fusion to intelligently integrate discriminative and robust features using early and late fusion strategies that allow DADNet to significantly consider the foreground and marginally utilize the background information. Furthermore, we impose diversity orthogonality regularization over several CNN layers, which boosts the performance of DADNet, resulting in an appropriate usage over IoT networks. A comprehensive set of ablation studies, comparison with other state-of-the-art approaches, and a time complexity analysis confirm the strength of our DADNet for re-identification tasks in AI-enabled IoT settings that are well-suited for a CPS.

Sport teams’ promotion of plant-based food consumption among fans

ABSTRACT The study’s goal is to examine the downstream relations of fan awareness of a favorite sport team’s pro-environmental food initiatives, considering fans with different degrees of team identification. Building upon the Awareness-to-Action Continuum, the authors propose two pathways that may help explain reported plant-based food consumption among fans (i.e. eating vegetables, eating vegan food, and avoiding meat), particularly at high (vs. low) degrees of team identification: team value internalization and fan-team personality match. Study 1 surveyed 799 fans of 12 different sport teams from various countries, which had implemented pro-environmental food initiatives in the past. Study 2 used a survey-experimental design to manipulate identification with one selected team from the U.S. (n = 356 fans) and assessed the downstream relations of fan awareness of the team’s pro-environmental food initiatives. The authors control for attitude and subjective norm in their analyses and show that the levels of team identification associate with the relation between fan awareness and reported consumption of plant-based foods via fan-team personality match but not via value internalization. Ironically, the perception of the team’s personality as superior to their own personality among low-identification fans is negatively related to the adoption of a plant-based diet despite high awareness levels.

The Cognition Mechanism and Intervention Effects of Listening to Senior’s Story: A Longitudinal Study of Learning Behaviors

<p>108課綱重視高中生的學習軌跡紀錄，引發對學習經驗效用之焦慮，影響學習表現。受前輩匹配觀啟發，採聽故事視角，觀察高中生接收大學前輩故事，於兩聽故事要素、兩心理知覺及三效標的貫時知覺變化。另檢驗在實驗組之學業成就t3、學習韌性t3、典範目標t3的知覺提升之前因與心理機轉，觀察思維學習t2、故事認同t2，對三效標影響，透過優勢自我t2、克服困難效能感t2的中介，觀察北區三所高中共135位，採重複量數、多元迴歸、拔靴法檢驗。顯示思維學習在前、後測表現，於實驗、對照組略具差異，故事認同則否，而優勢自我、克服困難效能感在跨三時點表現，兩組差異達顯著水準，且實驗組皆呈現上升趨勢，在學業成就、學習韌性、典範目標亦有鮮明差異。又檢驗實驗組之效果，對學習韌性t3，可由思維學習t2透過優勢自我t2造成影響。而典範目標t3，則由故事認同t2透過克服困難效能感t2造成影響，且思維學習t2亦有助益。對學業成就t3，由故事認同t2透過克服困難效能感t2造成影響，也能透過優勢自我t2的影響路徑。回饋理論，前輩匹配概念轉以聽故事角度來認識，能引發心理知覺及內在優勢自我及克服困難效能感的提升與保留，三類效標亦有提升。且三類效標有各自心理路徑與機轉。在諮商輔導工作上，可嘗試提供多元的前輩經驗，促使高中生有信心規劃學習歷程紀錄。</p> <p> </p><p>According to the new Curriculum Guidelines published in 108, high school students are required to submit their academic portfolios as part of the application package for college entrance, which has increased students’ anxiety about preparing academic portfolios and negatively influences their learning performance. This article, inspired by the perspective of Kuder’s personal match, used the research approach of story listening and evaluated the learning effects after students listened to seniors sharing their stories in high school. The effects of learning from seniors’ thoughts and identifying with seniors, which could impact students’ academic achievement, learning resilience, and role modeling, were evaluated by assessing the recognition of self-advantages and self-efficacy based on overcoming obstacles. An experimental longitudinal study was selected as the research design, and a sample of 218 students from three different high schools near Taipei was selected; repeated measures ANOVA and multiple regression were selected to examine the results. Comparing the pre- and post-test scores, the results showed that ""there was"" a significant difference between the experimental and comparison groups in the variable of learning from seniors’ thoughts, but no significant difference in the variable of identifying with seniors.. Meanwhile, there was a significant difference between the two groups in the variables of self-advantages and self-efficacy across the three time intervals; the performance of the experimental group showed an upward trend, while the performance of the comparison group showed a downward trend. In addition, learning resilience at t3 was influenced by the variable of learning from seniors’ thoughts at t2 through the mediation effects of recognizing self-advantages, and academic achievement at t3 was affected by learning from seniors’ thoughts at t2 via the mediation effects of self-efficacy from overcoming obstacles at t2. Furthermore, role modeling at t3 can be shaped by identifying seniors at t2 with the medication effects of self-efficacy from overcoming obstacles at t2. This shows that we can use the story listening approach to reframe the concept of personal matching, which can facilitate two psychological mechanisms that create two different effects. Similarly, it reveals that the intervention effects of story listening on the variable of learning from seniors’ thoughts are more substantial than identifying with seniors as well as enhancing the ability to recognize self-advantages and improving self-efficacy from overcoming obstacles longitudinally. In college counseling, counselors can provide different stories and experiences to help students prepare their academic portfolios with confidence and be successful in college applications.</p> <p> </p>

The Influences of Ad-Targeting on Teen Purchasing Decision Autonomy


 Due to an increased awareness and adherence to the social comparison theory, where individuals have a fundamental need to be liked and ranked amongst their peers, teens fall victim to the wishful identification phenomenon potentially leading to long-term dependence on materialism for fulfilment in life. Influencers derive their social powers from wishful identification because of their seeming relatability, often presenting themselves as ordinary and relatable the status of an influencer appears more commonly achievable than the fame of the typical, mainstream celebrity, microcelebrity audiences, “followers,” often demonstrate a greater desire to emulate and achieve the status of these “micro-celebrities,” in comparison to an A-list celebrity. Knowing this, the following study utilized a quantitative, quasi-experimental method to evaluate the effectiveness of targeted advertisements. To begin, the personality of an influencer was matched with the personality of a female teenage subject using the 5 Factors of Personality Test. After matches were made, everyone was presented with two influencer advertisements – one with the best personality match and mismatch – and an online survey evaluated whether the individual would buy the presenting product and why or why not, depending. Ad-targeting does prove to be successful in making impactful marketing matches for teen female subjects in influencer marketing; this boost is only noted where there is great enough personality resonation between the subject in the influencer in both introverted and extroverted traits to be able to establish the wishful identification connections needed to spur the persuasive forces of influencer marketing.

Dual-level contrastive learning for unsupervised person re-identification

Unsupervised person re-identification (re-ID) has drawn increasing attention in the community because it is more data-friendly when applied in the real world. Most existing works leverage instance discrimination learning to guide the model to learn image features for person matching. However, the instance discrimination may cause unexpected repulsion among similar samples, which makes the unsupervised feature learning unstable. To address this problem, we propose a dual-level contrastive learning (DLCL) framework to mine both the intra-instance and inter-instance similarities. The DLCL framework consists of two tasks: instance-instance contrastive learning (IICL) and instance-community contrastive learning (ICCL). The IICL aims to mine the intra-instance similarity via pulling an original sample and its augmented versions closer and pushing different instances away. The ICCL is proposed to capture the inter-instance similarity by attracting similar instances to the same sample community, which can reduce the unexpected repulsion brought by instance discrimination. The combination of IICL and ICCL can enable the model to learn more robust and discriminative image features. Extensive experimental results on Market-1501 and DukeMTMC-reID indicate the effectiveness of our method for unsupervised person re-ID.

603. Comparing assays for Clinical Phage Microbiology in biofilm

Abstract Background Non-resolving infections are commonly associated with bacterial biofilm formation. One promising solution is the use of lytic bacterial bacteriophages, as an adjunctive approach with antibiotics to penetrate biofilms and cure infections. A major challenge in the use of phages for therapy is the need for accurate personal matching of the most effective phage-antibiotic combination against the target bactera. This matching, termed “Clinical Phage Microbiology” (CPM) is even more problematic regarding biofilm and there are no standard methods for that. Here we compared several approaches for CPM in biofilms. Methods The efficacy of phages on the biofilm was tested in 8 methods using 5 phages at two concentrations. To this end five Pseudomonas aeruginosa (PA) specific phages were grown for biofilm establishment. Next, we followed the metabolic activity of the cultures using: (1)Agilent Seahorse XF Analyzers., and (2) Symcel’s calScreener calorimeter. In addition, we measured the levels of extracellular DNA in the biofilm supernatant using (3) rtPCR and (4) online by SYBR Green without amplification. The 5th method used GFP PA14 with an allure-red dye that allows the assessment of the adherent bacteria only. Last, we used the commonly used (6) Colony Forming Units (CFU) counts, (7) Crystal Violet (CV), and (8) Live/dead stain as our gold standard controls. Results Out of these methods, the calScreener (Figure 1A-F), and Seahorse were able to differentiate between phages real-time. The calSreener also showed a clear difference between biofilm and planktonic growth curves. The Real-Time PCR and the Live dead stain methods have shown a significant differentiation at the treatment endpoint. The other methods, CFU, CV, Syber Green, and Allura red did not show a significant difference between the phages. It should be noted that phages biofilm ranking did not correlate with the planktonic one. Figure 1 heat flow was measured using calScreener over 21 hours, comparing 5 different phages at a PFU of 108 and 106 on Pseudomonas aeruginosa planktonic (A) or biofilm (B-F). (A) planktonic bacteria and PASA16 phage treated, (B) KLEIN 3 phage, (C) PASA16 phage, (D) PB2F phage, (E) D9 phage, (F) PB2C phage. Figure 2 Comparison, using rtPCR, of 5 different phages at a PFU of 10^8 and 10^6 treated for 24 hours, and then their supernatant was filtered. Primers for housekeeping genes (A) ldhD, (B) rpoD. Conclusion We have compared 8 different methods as assays for phage screening. In this model of PA biofilm and phages, we found that calScreener and Seahorse are superior while CFU, CV, Syber Green, Allura red are not as useful. Disclosures Ran Nir-Paz, MD, BiomX: Advisor/Consultant|Technophage: Advisor/Consultant.

Superb fairy-wrens with extreme exploration phenotypes respond more strongly to simulated territory intrusions

Individuals can exhibit strong consistencies in behaviours across time and contexts, referred to as animal personalities. Yet studies linking personality differences measured in captivity to ecologically relevant social behaviours in the wild are sparse and contradictory. Here, we used wild superb fairy-wrens, Malurus cyaneus , to investigate whether two personality traits measured in short-term captivity (exploration and aggressiveness) predicted aggressiveness in the wild (measured as territory defence). We first identified behavioural phenotypes within our study population by temporarily capturing adult birds to measure their (1) emergence, activity and sector visits in a novel environment (exploration) and (2) aggression in response to mirror image stimulation (aggressiveness). We later simulated two conspecific intrusions within their territory using playback of vocalizations (female or male chatter song) within their territory, in a pairwise design. Both exploration and breeding status (breeder/dominant versus helper/subordinate) influenced the birds’ response to the simulated intrusions. Dominant individuals and individuals with extreme exploration phenotypes (i.e. fast or slow explorers) responded more strongly to playback compared to helpers or those with intermediate exploration phenotypes, respectively. However, aggressiveness measured in short-term captivity (mirror response) did not correlate with our measure of aggressiveness in the wild (playback response during territory defence). The personality matching of breeding pairs (whether their partner had a matching or mismatching exploration phenotype) also influenced playback response, as pairs with mismatching exploration phenotypes responded more strongly to conspecific intruders than pairs with similar phenotypes. Our results support a growing number of studies showing the importance of animal personalities for response strategies and social behaviours. • We explored associations between personality traits in wild superb fairy-wrens. • Birds with extreme exploration phenotypes were more aggressive against conspecifics. • Aggressiveness in captivity did not correlate with aggressiveness in the wild. • Pairs with mismatching exploration phenotypes responded more strongly to intruders. • Our results support the validity of personality in fitness-related contexts.

Illumination Unification for Person Re-Identification

The performance of person re-identification (re-ID) is easily affected by illumination variations caused by different shooting times, places and cameras. Existing illumination-adaptive methods usually require annotating cross-camera pedestrians on each illumination scale, which is unaffordable for a long-term person retrieval system. The cross-illumination person retrieval problem presents a great challenge for accurate person matching. In this paper, we propose a novel method to tackle this task, which only needs to annotate pedestrians on one illumination scale. Specifically, (i) we propose a novel Illumination Estimation and Restoring framework (IER) to estimate the illumination scale of testing images taken at different illumination conditions and restore them to the illumination scale of training images, such that the disparities between training images with uniform illumination and testing images with varying illuminations are reduced. IER achieves promising results on illumination-adaptive dataset and proving itself a proper baseline for cross-illumination person re-ID. (ii) we propose a Mixed Training strategy using both Original and Reconstructed images (MTOR) to further improve model performance. We generate reconstructed images that are consistent with the original training images in content but more similar to the restored images in style. The reconstructed images are combined with the original training images for supervised training to further reduce the domain gap between original training images and restored testing images. To verify the effectiveness of our method, some simulated illumination-adaptive datasets are constructed with various illumination conditions. Extensive experimental results on the simulated datasets validate the effectiveness of the proposed method. The source code is available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/FadeOrigin/IUReId</uri> .

2021 Census to Census Coverage Survey Matching Results.

The 2021 England and Wales Census was matched to the Census Coverage Survey (CCS). This was an essential requisite for estimating undercount in the Census. To ensure outputs could be produced within a year of census day, matching had to be completed in eight weeks.&#x0D; We used the Data Access Platform to write efficient and accurate automatic matching algorithms. Probabilistic and associative matching methods were used to present candidate pairs that could not be automatically matched but likely consisted of many true matches. These were resolved by matchers in the clerical matching system. Active learning was used to iteratively improve probabilistic parameters as matches were made.&#x0D; We also developed a pre-search algorithm that simplifies clerical matching by replacing clerical searching (here’s a record, can you find a match?) with clerical resolution (here are two records, do they match?).&#x0D; As a result of our improvements, we increased our automatic matching rates from 70% (2011 Census) to 93% (2021 Census) for person matching and from 60% to 95% for household matching, without loss of accuracy. Precision and recall were estimated to both be 99.96% for person matching and 100% and 99.78% respectively for household matching. For both persons and households, the same trends seen in 2011 are seen here too; however, most of the biases have decreased and most match rates have increased.&#x0D; Dual system estimation using the matched and unmatched records enabled us to estimate that coverage in the 2021 England and Wales Census was 97%.&#x0D; The matching methodology implemented for the 2021 Census to CCS matching improved both person and household matching greatly when compared with the 2011 matching. The clerical matching system enabled clerical matching at scale and at pace. Thus, we successfully completed all clerical matching within our eight-week deadline.

Hybrid Modality Metric Learning for Visible-Infrared Person Re-Identification

Visible-infrared person re-identification (Re-ID) has received increasing research attention for its great practical value in night-time surveillance scenarios. Due to the large variations in person pose, viewpoint, and occlusion in the same modality, as well as the domain gap brought by heterogeneous modality, this hybrid modality person matching task is quite challenging. Different from the metric learning methods for visible person re-ID, which only pose similarity constraints on class level, an efficient metric learning approach for visible-infrared person Re-ID should take both the class-level and modality-level similarity constraints into full consideration to learn sufficiently discriminative and robust features. In this article, the hybrid modality is divided into two types, within modality and cross modality. We first fully explore the variations that hinder the ranking results of visible-infrared person re-ID and roughly summarize them into three types: within-modality variation, cross-modality modality-related variation, and cross-modality modality-unrelated variation. Then, we propose a comprehensive metric learning framework based on four kinds of paired-based similarity constraints to address all the variations within and cross modality. This framework focuses on both class-level and modality-level similarity relationships between person images. Furthermore, we demonstrate the compatibility of our framework with any paired-based loss functions by giving detailed implementation of combing it with triplet loss and contrastive loss separately. Finally, extensive experiments of our approach on SYSU-MM01 and RegDB demonstrate the effectiveness and superiority of our proposed metric learning framework for visible-infrared person Re-ID.

Comparing Special Education Teachers’ Personality Profile With Their Choice to Teach

Researchers have yet to examine the association of Holland personality profiles as it relates to special education teachers. In response to this need, we report the personality and vocational profiles (Holland Codes) of 134 special education teachers across a special education training program. The purpose of this paper is to summarize findings from the Self-Directed Search measure commonly used to assess the personality of participants in an occupation and suggest implications for participants’ choice in becoming a special education teacher. Our focus was on personality match with vocational choice to include participants’ demographic (e.g., gender, race/ethnicity, and geographical location) profile. Findings from the study reveal that while special educators’ overall personality profile is congruent with the Holland Codes associated with special education teachers, other features may explain participants’ choice to pursue a career as a special education teacher. Implications for teacher preparation programs and K-12 schools training are recommended based on the research findings.

Learning hybrid ranking representation for person re-identification

Contemporary person re-identification (re-id) methods mostly compute independentlya feature representation of each person image in the query set and the gallery set. This strategy fails to consider any ranking context information of each probe image in the query set represented implicitly by the whole gallery set. Some recent re-ranking re-id methods therefore propose to take a post-processing strategy to exploit such contextual information for improving re-id matching performance. However, post-processing is independent of model training without jointly optimising the re-id feature and the ranking context information for better compatibility. In this work, for the first time, we show that the appearance feature and the ranking context information can be jointly optimised for learning more discriminative representations and achieving superior matching accuracy. Specifically, we propose to learn a hybrid ranking representation for person re-id with a two-stream architecture: (1) In the external stream, we use the ranking list of each probe image to learn plausible visual variations among the top ranks from the gallery as the external ranking information; (2) In the internal stream, we employ the part-based fine-grained feature as the internal ranking information, which mitigates the harm of incorrect matches in the ranking list. Assembling these two streams generates a hybrid ranking representation for person matching. Extensive experiments demonstrate the superiority of our method over the state-of-the-art methods on four large-scale re-id benchmarks (Market-1501, DukeMTMC-ReID, CUHK03 and MSMT17), under both supervised and unsupervised settings.

Why do consumers with social phobia prefer anthropomorphic customer service chatbots? Evolutionary explanations of the moderating roles of social phobia

Drawing from evolutionary psychology of anthropomorphism and social phobia, two between-subjects experiments examined the effects of different types of customer service chatbots. Experiment 1 supports the interaction effects between chatbots’ anthropomorphism and consumers’ social phobia on continuance use intention and willingness to recommend the chatbot. Consumers with high social phobia prefer anthropomorphic chatbots to less anthropomorphic chatbots. Experiment 2 confirms the moderating role of social phobia in determining the effects of consumer-chatbot personality matching (similarity attraction) vs. mismatching (complementarity attraction) on the outcome variables only for competent chatbots. For the consumer-chatbot personality mismatching condition, developing competent chatbots for less conscientious consumers with high social phobia will help alleviate socially isolated consumers’ social pain, while developing too smart chatbots for less conscientious consumers with low social phobia can have detrimental effects. Evolutionary psychological mechanisms and managerial implications for chatbot developers and creative directors are explained.

Sequential End-to-end Network for Efficient Person Search

Person search aims at jointly solving Person Detection and Person Re-identification (re-ID). Existing works have designed end-to-end networks based on Faster R-CNN. However, due to the parallel structure of Faster R-CNN, the extracted features come from the low-quality proposals generated by the Region Proposal Network, rather than the detected high-quality bounding boxes. Person search is a fine-grained task and such inferior features will significantly reduce re-ID performance. To address this issue, we propose a Sequential End-to-end Network (SeqNet) to extract superior features. In SeqNet, detection and re-ID are considered as a progressive process and tackled with two sub-networks sequentially. In addition, we design a robust Context Bipartite Graph Matching (CBGM) algorithm to effectively employ context information as an important complementary cue for person matching. Extensive experiments on two widely used person search benchmarks, CUHK-SYSU and PRW, have shown that our method achieves state-of-the-art results. Also, our model runs at 11.5 fps on a single GPU and can be integrated into the existing end-to-end framework easily.

Beyond modality alignment: Learning part-level representation for visible-infrared person re-identification

Visible-Infrared person re-IDentification (VI-reID) aims to automatically retrieve the pedestrian of interest exposed to sensors in different modalities, such as visible camera v.s. infrared sensor. It struggles to learn both modality-invariant and discriminant representations. Unfortunately, existing VI-reID work mainly focuses on tackling the modality difference, which fine-grained level discriminant information has not been well investigated. This causes inferior identification performance. To address the problem, we propose a Dual-Alignment Part-aware Representation (DAPR) framework to simultaneously alleviate the modality bias and mine different level of discriminant representations. Particularly, our DAPR reduces modality discrepancy of high-level features hierarchically by back-propagating reversal gradients from a modality classifier, in order to learn a modality-invariant feature space. And meanwhile, multiple heads of classifiers with the improved part-aware BNNeck are integrated to supervise the network producing identity-discriminant representations w.r.t. both local details and global structures in the learned modality-invariant space. By training in an end-to-end manner, the proposed DAPR produces camera-modality-invariant yet discriminant features11The implementation code is available on https://github.com/pzhang0610/DAPR for the purpose of person matching across modalities. Extensive experiments are conducted on two benchmarks, i.e., SYSU MM01 and RegDB, and the results demonstrate the effectiveness of our proposed method.