pith. sign in

arxiv: 2606.30458 · v1 · pith:PUXXF5JJnew · submitted 2026-06-29 · 💻 cs.CV

Cross-Resolution Semantic Transfer for Robust Text-to-Image Retrieval in Low-Resolution Surveillance

Wenjie Qian , Bin Yang , Xiao Wang , Wenke Huang , Ling Mei , Xin Xu , Mang Ye This is my paper

Pith reviewed 2026-06-30 06:49 UTC · model grok-4.3

classification 💻 cs.CV
keywords text-to-image retrievalperson re-identificationcross-resolutionlow-resolution surveillancesemantic transferCLIPresolution-conditioned reasoning
0
0 comments X

The pith

CRST transfers semantic information across resolutions to fix reliability collapse and ranking drift in text-to-image person retrieval.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper identifies two failure modes in text-to-image person re-identification under real-world resolution variance: evidence reliability collapse where low-res visual tokens lose grounding power, and ranking distribution drift where mixed galleries destabilize similarity rankings. It introduces the Cross-Resolution Semantic Transfer framework with three modules to estimate token reliability, recover cues through text guidance, and transfer high-resolution neighborhood structure to low-resolution cases. A reader would care because surveillance systems routinely mix high- and low-resolution footage, and current CLIP-based methods degrade sharply on the low end. The method reports average gains of 5.7 percent Rank-1 and 5.3 percent mAP on ultra-low-resolution splits of three benchmarks while leaving high-resolution accuracy unchanged.

Core claim

The central claim is that the CRST CLIP-style framework, built from resolution-conditioned reasoning to suppress unreliable tokens, text-guided refinement to inject semantic priors, and CR-RDA to transfer HR neighborhood geometry, mitigates evidence reliability collapse and ranking distribution drift, delivering 5.7 percent and 5.3 percent average gains in ultra-low-resolution Rank-1 and mAP on CUHK-PEDES, ICFG-PEDES, and RSTPReid while stabilizing mixed-resolution retrieval without loss on high-resolution data.

What carries the argument

Cross-Resolution Semantic Transfer (CRST) framework using resolution-conditioned reasoning, text-guided refinement, and CR-RDA to move semantic structure from high-resolution to low-resolution inputs.

If this is right

  • Ultra-low-resolution Rank-1 rises 5.7 percent and mAP rises 5.3 percent on average across the three person re-identification benchmarks.
  • Mixed-resolution galleries produce stable similarity rankings instead of distorted neighborhoods.
  • High-resolution retrieval accuracy stays the same.
  • The approach works inside existing CLIP-style text-to-image pipelines for surveillance.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same resolution-conditioned token weighting could be tested on video retrieval or multi-camera tracking where frame quality also varies.
  • Combining CRST with separate super-resolution preprocessing might produce additive gains on the lowest-resolution inputs.
  • Deployment on live camera feeds would reveal whether the neighborhood transfer remains stable under streaming resolution changes.

Load-bearing premise

The three modules actually correct evidence reliability collapse and ranking distribution drift on data outside the three evaluation sets rather than merely fitting those sets.

What would settle it

Running CRST on a new surveillance dataset with previously unseen resolution mixtures and finding no gain in ultra-low-resolution Rank-1 or persistent ranking instability after applying the modules would falsify the claim.

Figures

Figures reproduced from arXiv: 2606.30458 by Bin Yang, Ling Mei, Mang Ye, Wenjie Qian, Wenke Huang, Xiao Wang, Xin Xu.

Figure 1
Figure 1. Figure 1: Illustration of motivation. Problem I) Evidence Reli￾ability Collapse (ERC): Resolution degradation corrupts fine￾grained evidence, causing cross-modal mismatch and incor￾rect top-ranked retrieval. Problem II) Ranking Distribution Drift (RDD): Mixed-resolution galleries distort similarity or￾dering, leading to ranking inconsistency across resolutions. settings, where the retrieval space is implicitly assum… view at source ↗
Figure 2
Figure 2. Figure 2: The pipeline of the CRST. CRST mitigates ERC and RDD by enforcing HR-referenced robustness constraints on [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Progressive degradation. We compare baseline and CRST across four resolution settings, reporting Rank-1 and mAP. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Training dynamics under cross-resolution train [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Text-guided semantic recovery under UltraLR. Left [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 8
Figure 8. Figure 8: Visualization of attention maps. (a) Input images, [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
read the original abstract

Text-to-image person re-identification (TIPR) retrieves target persons using natural language descriptions. However, existing methods largely overlook resolution variance in real-world surveillance. They characterize cross-resolution TIPR through two coupled failure modes: Evidence Reliability Collapse (ERC), where degraded visual tokens become unreliable for grounding fine-grained text, and Ranking Distribution Drift (RDD), where mixed-resolution galleries distort similarity neighborhoods and destabilize retrieval rankings. To address this challenge, we propose Cross-Resolution Semantic Transfer (CRST), a CLIP-style framework with three modules: resolution-conditioned reasoning, text-guided refinement and CR-RDA. Resolution-conditioned reasoning estimates token reliability to suppress corrupted evidence. Text-guided refinement injects semantic priors to recover discriminative cues. CR-RDA transfers HR neighborhood geometry to stabilize LR ranking under mixed resolutions. Experiments on CUHK-PEDES, ICFG-PEDES, and RSTPReid show that CRST improves ultra-low-resolution Rank-1 and mAP on average by 5.7% and 5.3%, while stabilizing mixed-resolution retrieval without sacrificing high-resolution accuracy.The code will be made publicly available.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The paper introduces Cross-Resolution Semantic Transfer (CRST), a CLIP-style framework for text-to-image person re-identification that targets resolution variance in surveillance imagery. It defines two coupled failure modes—Evidence Reliability Collapse (ERC) and Ranking Distribution Drift (RDD)—and proposes three modules (resolution-conditioned reasoning to estimate token reliability, text-guided refinement to inject semantic priors, and CR-RDA to transfer HR neighborhood geometry) to mitigate them. Experiments on CUHK-PEDES, ICFG-PEDES, and RSTPReid report average gains of 5.7% Rank-1 and 5.3% mAP in ultra-low-resolution settings while preserving high-resolution accuracy and stabilizing mixed-resolution retrieval; code release is promised.

Significance. If the modules demonstrably reduce ERC and RDD on unseen data rather than fitting the three evaluation sets, the work would provide a practical advance for real-world TIPR under variable surveillance resolutions. The explicit modeling of token reliability and cross-resolution neighborhood transfer is a targeted contribution, and the promised public code supports reproducibility.

major comments (2)
  1. [Experiments] Experiments section: All quantitative results are confined to CUHK-PEDES, ICFG-PEDES, and RSTPReid with no cross-dataset transfer, held-out surveillance collection, or explicit ERC/RDD metrics (e.g., token reliability scores or ranking stability measures). This leaves the central claim—that the three modules suppress ERC and RDD rather than capitalize on dataset idiosyncrasies—unverified and load-bearing for the reported 5.7%/5.3% gains.
  2. [Abstract] Abstract and Methods: No error bars, ablation controls, or dataset statistics are referenced, and the abstract provides no derivation details for how resolution-conditioned reasoning or CR-RDA are implemented or optimized. Without these, the soundness of the empirical gains cannot be assessed.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the thoughtful and constructive report. We address the major comments point-by-point below, indicating where revisions will be made to strengthen the manuscript while clarifying aspects already present in the work.

read point-by-point responses

  1. Referee: [Experiments] Experiments section: All quantitative results are confined to CUHK-PEDES, ICFG-PEDES, and RSTPReid with no cross-dataset transfer, held-out surveillance collection, or explicit ERC/RDD metrics (e.g., token reliability scores or ranking stability measures). This leaves the central claim—that the three modules suppress ERC and RDD rather than capitalize on dataset idiosyncrasies—unverified and load-bearing for the reported 5.7%/5.3% gains.

    Authors: The three datasets are the standard benchmarks for text-to-image person re-identification and already span multiple surveillance scenarios with varying resolution characteristics. Consistent gains across all three support that the improvements are not dataset-specific. We agree, however, that direct quantification of ERC and RDD would make the mechanistic claims more verifiable. In the revision we will add explicit metrics (token reliability scores before/after resolution-conditioned reasoning and ranking stability measures before/after CR-RDA) to the experiments section. Cross-dataset transfer and new held-out collections lie outside the current experimental scope; we will note this limitation explicitly. revision: partial

  2. Referee: [Abstract] Abstract and Methods: No error bars, ablation controls, or dataset statistics are referenced, and the abstract provides no derivation details for how resolution-conditioned reasoning or CR-RDA are implemented or optimized. Without these, the soundness of the empirical gains cannot be assessed.

    Authors: The abstract is a high-level summary; the methods section already contains the full mathematical formulations, optimization objectives, and architectural details for resolution-conditioned reasoning and CR-RDA. Ablation tables demonstrating module contributions are present in the experiments. We will add error bars to all reported results, include dataset statistics (image counts, resolution distributions), and ensure the abstract briefly references the core technical approach. These changes will be incorporated in the revised manuscript. revision: yes

standing simulated objections not resolved
  • Addition of new held-out surveillance collections or full cross-dataset transfer experiments, which would require substantial new data acquisition and compute beyond the revision timeline.

Circularity Check

0 steps flagged

No circularity; derivation self-contained with no reductions to inputs

full rationale

The provided abstract and description introduce CRST as a CLIP-style framework with three additive modules (resolution-conditioned reasoning, text-guided refinement, CR-RDA) to mitigate ERC and RDD failure modes, reporting empirical gains on three standard datasets. No equations, fitted parameters renamed as predictions, self-definitional constructs, or load-bearing self-citations appear in the text. The central claims rest on experimental results rather than any derivation that reduces by construction to its own inputs or prior author work.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only abstract available; no free parameters, axioms, or invented entities can be extracted.

pith-pipeline@v0.9.1-grok · 5741 in / 1006 out tokens · 24269 ms · 2026-06-30T06:49:59.518976+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

63 extracted references · 8 canonical work pages · 2 internal anchors

  1. [1]

    Kai Niu, Yanyi Liu, Yuzhou Long, Yan Huang, Liang Wang, and Yanning Zhang

  2. [2]

    An overview of text-based person search: recent advances and future directions.IEEE Transactions on Circuits and Systems for Video Technology, 34, 9, 7803–7819

  3. [3]

    Jialong Zuo, Hanyu Zhou, Ying Nie, Feng Zhang, Tianyu Guo, Nong Sang, Yunhe Wang, and Changxin Gao. 2024. Ufinebench: towards text-based person retrieval with ultra-fine granularity. (2024). https://arxiv.org/abs/2312.03441 arXiv: 2312.03441[cs.CV]

    work page arXiv 2024

  4. [4]

    Yukang Zhang and Hanzi Wang. 2023. Diverse embedding expansion net- work and low-light cross-modality benchmark for visible-infrared person re- identification. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2153–2162

    2023

  5. [5]

    Mehdi Cherti, Romain Beaumont, Ross Wightman, Mitchell Wortsman, Gabriel Ilharco, Cade Gordon, Christoph Schuhmann, Ludwig Schmidt, and Jenia Jitsev

  6. [6]

    InPro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2818–2829

    Reproducible scaling laws for contrastive language-image learning. InPro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2818–2829

  7. [9]

    Byoungjip Kim, Sungik Choi, Dasol Hwang, Moontae Lee, and Honglak Lee

  8. [10]

    Transferring pre-trained multimodal representations with cross-modal similarity matching.Advances in Neural Information Processing Systems, 35, 30826–30839

  9. [11]

    Xiaohan Xu, Ming Li, Chongyang Tao, Tao Shen, Reynold Cheng, Jinyang Li, Can Xu, Dacheng Tao, and Tianyi Zhou. 2024. A survey on knowledge distillation of large language models.arXiv preprint arXiv:2402.13116

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  10. [12]

    Zongzong Wu, Xiangchun Yu, Donglin Zhu, Qingwei Pang, Shitao Shen, Teng Ma, and Jian Zheng. 2022. Sr-dsff and fenet-reid: a two-stage approach for cross resolution person re-identification.Computational Intelligence and Neuroscience, 2022, 1, 4398727

    2022

  11. [13]

    Shuanglin Yan, Jun Liu, Neng Dong, Liyan Zhang, and Jinhui Tang. 2025. Cross- modal collaborative representation learning for text-to-image person retrieval. InProceedings of the thirty-fourth international joint conference on artificial intelligence (IJCAI-25)

    2025

  12. [14]

    Bin Yang, Jun Chen, Cuiqun Chen, and Mang Ye. 2023. Dual consistency- constrained learning for unsupervised visible-infrared person re-identification. IEEE Transactions on Information Forensics and Security

    2023

  13. [15]

    Wentan Tan, Changxing Ding, Jiayu Jiang, Fei Wang, Yibing Zhan, and Dapeng Tao. 2024. Harnessing the power of mllms for transferable text-to-image person reid. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 17127–17137

    2024

  14. [16]

    Junnan Li, Dongxu Li, Silvio Savarese, and Steven Hoi. 2023. Blip-2: boot- strapping language-image pre-training with frozen image encoders and large language models. InInternational conference on machine learning. PMLR, 19730– 19742

    2023

  15. [17]

    Zhangyi Hu, Bin Yang, and Mang Ye. [n. d.] Empowering visible-infrared person re-identification with large foundation models. InThe Thirty-eighth Annual Conference on Neural Information Processing Systems

  16. [18]

    Bin Yang, Jun Chen, Xianzheng Ma, and Mang Ye. 2023. Translation, associa- tion and augmentation: learning cross-modality re-identification from single- modality annotation.IEEE Transactions on Image Processing

    2023

  17. [19]

    Lin Yuanbo Wu, Lingqiao Liu, Yang Wang, Zheng Zhang, Farid Boussaid, Mohammed Bennamoun, and Xianghua Xie. 2023. Learning resolution-adaptive representations for cross-resolution person re-identification.IEEE Transactions on Image Processing, 32, 4800–4811

    2023

  18. [20]

    Yukang Zhang, Yan Yan, Yang Lu, and Hanzi Wang. 2021. Towards a uni- fied middle modality learning for visible-infrared person re-identification. In Proceedings of the 29th ACM International Conference on Multimedia, 788–796

    2021

  19. [21]

    Delong Liu, Haiwen Li, Zhicheng Zhao, and Yuan Dong. 2025. Text-guided image restoration and semantic enhancement for text-to-image person retrieval. Neural Networks, 184, 107028

    2025

  20. [22]

    Zhiqi Pang, Lingling Zhao, and Chunyu Wang. 2024. Dual-resolution fusion modeling for unsupervised cross-resolution person re-identification. InPro- ceedings of the 32nd ACM international conference on multimedia, 4063–4072

    2024

  21. [23]

    Yukang Zhang, Yan Yan, Jie Li, and Hanzi Wang. 2023. Mrcn: a novel modality restitution and compensation network for visible-infrared person re-identification. InProceedings of the AAAI Conference on Artificial Intelligencenumber 3. Vol. 37, 3498–3506

    2023

  22. [24]

    Fanzhi Jiang, Su Yang, Mark W Jones, and Liumei Zhang. 2025. From at- tributes to natural language: a survey and foresight on text-based person re-identification.Information Fusion, 118, 102879

    2025

  23. [25]

    Bin Yang, Jun Chen, and Mang Ye. 2023. Towards grand unified representation learning for unsupervised visible-infrared person re-identification. InProceed- ings of the IEEE/CVF International Conference on Computer Vision (ICCV). (Oct. 2023), 11069–11079

    2023

  24. [26]

    Bin Yang, Mang Ye, Jun Chen, and Zesen Wu. 2022. Augmented dual-contrastive aggregation learning for unsupervised visible-infrared person re-identification. InACM MM, 2843–2851

    2022

  25. [27]

    Yunyao Mao, Wengang Zhou, Zhenbo Lu, Jiajun Deng, and Houqiang Li. 2022. Cmd: self-supervised 3d action representation learning with cross-modal mu- tual distillation. InEuropean Conference on Computer Vision. Springer, 734– 752

    2022

  26. [28]

    Zhenyu Cui, Jiahuan Zhou, and Yuxin Peng. 2024. Dma: dual modality-aware alignment for visible-infrared person re-identification.IEEE Transactions on Information Forensics and Security, 19, 2696–2708

    2024

  27. [29]

    Alexey Dosovitskiy. 2020. An image is worth 16x16 words: transformers for image recognition at scale.arXiv preprint arXiv:2010.11929

    work page internal anchor Pith review Pith/arXiv arXiv 2020

  28. [30]

    Chuanguang Yang, Zhulin An, Libo Huang, Junyu Bi, Xinqiang Yu, Han Yang, Boyu Diao, and Yongjun Xu. 2024. Clip-kd: an empirical study of clip model distillation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 15952–15962

    2024

  29. [31]

    Xiao Wang, Lekai Liu, Bin Yang, Mang Ye, Zheng Wang, and Xin Xu. 2025. To- kenmatcher: diverse tokens matching for unsupervised visible-infrared person re-identification. InProceedings of the AAAI Conference on Artificial Intelligence number 8. Vol. 39, 7934–7942

    2025

  30. [32]

    Bin Yang, Jun Chen, and Mang Ye. 2024. Shallow-deep collaborative learning for unsupervised visible-infrared person re-identification. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 16870–16879

    2024

  31. [33]

    Yukang Zhang, Xinwen Fan, Yujun Yang, Yang Lu, and Hanzi Wang. 2025. Image-attribute and frequency-spatial dual collaborative learning for pedes- trian attribute recognition.IEEE Transactions on Information Forensics and Security, 20, 11715–11727

    2025

  32. [34]

    Kanchana Vaishnavi Gandikota and Paramanand Chandramouli. 2024. Text- guided explorable image super-resolution. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, 25900–25911

    2024

  33. [35]

    Youbo Lei, Feifei He, Chen Chen, Yingbin Mo, Sijia Li, Defeng Xie, and Haonan Lu. 2024. Mcad: multi-teacher cross-modal alignment distillation for efficient image-text retrieval. InFindings of the Association for Computational Linguistics: NAACL 2024, 1491–1503

    2024

  34. [36]

    Shuping Hui, Min Wang, Hui Wu, Wengang Zhou, and Houqiang Li. 2025. Csd: cross-modal similarity distillation for zero-shot composed image retrieval. In Proceedings of the 6th Workshop on Intelligent Cross-Data Analysis and Retrieval, 1–8

    2025

  35. [37]

    Shuanglin Yan, Neng Dong, Liyan Zhang, and Jinhui Tang. 2023. Clip-driven fine-grained text-image person re-identification.IEEE Transactions on Image Processing, 32, 6032–6046

    2023

  36. [38]

    Zhiyin Shao, Xinyu Zhang, Changxing Ding, Jian Wang, and Jingdong Wang

  37. [39]

    InProceedings of the IEEE/CVF international conference on com- puter vision, 11174–11184

    Unified pre-training with pseudo texts for text-to-image person re- identification. InProceedings of the IEEE/CVF international conference on com- puter vision, 11174–11184

  38. [40]

    Min Cao, Yang Bai, Ziyin Zeng, Mang Ye, and Min Zhang. 2024. An empirical study of clip for text-based person search. InProceedings of the AAAI Conference on Artificial Intelligence. Vol. 38, 465–473

    2024

  39. [41]

    Zehong Ma, Hao Chen, Wei Zeng, Limin Su, and Shiliang Zhang. 2025. Multi- modal reference learning for fine-grained text-to-image retrieval.IEEE Trans- actions on Multimedia

    2025

  40. [42]

    Yating Liu, Zimo Liu, Xiangyuan Lan, Wenming Yang, Yaowei Li, and Qingmin Liao. 2025. Dm-adapter: domain-aware mixture-of-adapters for text-based person retrieval. InProceedings of the AAAI Conference on Artificial Intelligence number 6. Vol. 39, 5703–5711. Conference’17, July 2017, Washington, DC, USA Trovato et al

    2025

  41. [43]

    Ding Jiang and Mang Ye. 2023. Cross-modal implicit relation reasoning and aligning for text-to-image person retrieval. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2787–2797

    2023

  42. [44]

    Ying Zhang and Huchuan Lu. 2018. Deep cross-modal projection learning for image-text matching. InProceedings of the European conference on computer vision (ECCV), 686–701

    2018

  43. [45]

    Nikolaos Sarafianos, Xiang Xu, and Ioannis A Kakadiaris. 2019. Adversarial rep- resentation learning for text-to-image matching. InProceedings of the IEEE/CVF international conference on computer vision, 5814–5824

    2019

  44. [46]

    Chenyang Gao, Guanyu Cai, Xinyang Jiang, Feng Zheng, Jun Zhang, Yifei Gong, Pai Peng, Xiaowei Guo, and Xing Sun. 2021. Contextual non-local alignment over full-scale representation for text-based person search.arXiv preprint arXiv:2101.03036

    work page arXiv 2021

  45. [47]

    Zijie Wang, Aichun Zhu, Jingyi Xue, Xili Wan, Chao Liu, Tian Wang, and Yifeng Li. 2022. Look before you leap: improving text-based person retrieval by learning a consistent cross-modal common manifold. InProceedings of the 30th ACM international conference on multimedia, 1984–1992

    2022

  46. [48]

    Xiujun Shu, Wei Wen, Haoqian Wu, Keyu Chen, Yiran Song, Ruizhi Qiao, Bo Ren, and Xiao Wang. 2022. See finer, see more: implicit modality alignment for text-based person retrieval. InEuropean Conference on Computer Vision. Springer, 624–641

    2022

  47. [49]

    Ammarah Farooq, Muhammad Awais, Josef Kittler, and Syed Safwan Khalid

  48. [50]

    InProceedings of the AAAI conference on artificial intelligence

    Axm-net: implicit cross-modal feature alignment for person re-identification. InProceedings of the AAAI conference on artificial intelligence. Vol. 36, 4477– 4485

  49. [51]

    Shuanglin Yan, Hao Tang, Liyan Zhang, and Jinhui Tang. 2023. Image-specific information suppression and implicit local alignment for text-based person search.IEEE transactions on neural networks and learning systems

    2023

  50. [52]

    Zefeng Ding, Changxing Ding, Zhiyin Shao, and Dacheng Tao. 2021. Semanti- cally self-aligned network for text-to-image part-aware person re-identification. arXiv preprint arXiv:2107.12666

    work page arXiv 2021

  51. [53]

    Shuanglin Yan, Neng Dong, Jun Liu, Liyan Zhang, and Jinhui Tang. 2023. Learning comprehensive representations with richer self for text-to-image person re-identification. InProceedings of the 31st ACM international conference on multimedia, 6202–6211

    2023

  52. [54]

    Fei Shen, Xiangbo Shu, Xiaoyu Du, and Jinhui Tang. 2023. Pedestrian-specific bipartite-aware similarity learning for text-based person retrieval. InProceed- ings of the 31st ACM International Conference on Multimedia, 8922–8931

    2023

  53. [55]

    Guanshuo Wang, Fufu Yu, Junjie Li, Qiong Jia, and Shouhong Ding. 2023. Exploiting the textual potential from vision-language pre-training for text- based person search.arXiv preprint arXiv:2303.04497

    work page arXiv 2023

  54. [56]

    Weihao Li, Lei Tan, Pingyang Dai, and Yan Zhang. 2024. Prompt decoupling for text-to-image person re-identification.arXiv preprint arXiv:2401.02173

    work page arXiv 2024

  55. [57]

    Yang Qin, Yingke Chen, Dezhong Peng, Xi Peng, Joey Tianyi Zhou, and Peng Hu

  56. [58]

    InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 27197–27206

    Noisy-correspondence learning for text-to-image person re-identification. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 27197–27206

  57. [59]

    Shuang Li, Tong Xiao, Hongsheng Li, Bolei Zhou, Dayu Yue, and Xiaogang Wang. 2017. Person search with natural language description. InProceedings of the IEEE conference on computer vision and pattern recognition, 1970–1979

    2017

  58. [60]

    Liang Zheng, Liyue Shen, Lu Tian, Shengjin Wang, Jingdong Wang, and Qi Tian. 2015. Scalable person re-identification: a benchmark. InProceedings of the IEEE international conference on computer vision, 1116–1124

    2015

  59. [61]

    Xiying Zheng, Yukang Zhang, Yang Lu, and Hanzi Wang. 2024. Semi-supervised visible-infrared person re-identification via modality unification and confi- dence guidance. InProceedings of the 32nd ACM International Conference on Multimedia, 5761–5770

    2024

  60. [62]

    Yukang Zhang, Yan Yan, Yang Lu, and Hanzi Wang. 2024. Adaptive middle modality alignment learning for visible-infrared person re-identification.Inter- national Journal of Computer Vision, 1–21

    2024

  61. [63]

    Mang Ye, Jianbing Shen, Gaojie Lin, Tao Xiang, Ling Shao, and Steven CH Hoi

  62. [64]

    Deep learning for person re-identification: a survey and outlook.IEEE transactions on pattern analysis and machine intelligence, 44, 6, 2872–2893

  63. [65]

    Yukang Zhang, Yang Lu, Yan Yan, Hanzi Wang, and Xuelong Li. 2024. Frequency domain nuances mining for visible-infrared person re-identification.arXiv preprint arXiv:2401.02162. Received 20 February 2007; revised 12 March 2009; accepted 5 June 2009

    work page arXiv 2024