Pith. sign in

REVIEW 2 major objections 6 minor 35 references

A hierarchical hyperbolic embedding keeps both cross-view identity consistency and view-specific cues for aerial-ground person re-identification.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-13 04:49 UTC pith:CNM5OB4J

load-bearing objection Solid AG-ReID methods paper: hierarchical Lorentz parent/child embeddings plus text-guided multi-layer fusion give consistent mAP gains; geometry is a design choice, not a proof. the 2 major comments →

arxiv 2607.09186 v1 pith:CNM5OB4J submitted 2026-07-10 cs.CV

HiHR: Hierarchical Hyperbolic Representation for Aerial-Ground Person Re-Identification

Qiwei Yang , Pingping Zhang This is my paper
classification cs.CV
keywords Aerial-Ground Person Re-IdentificationHyperbolic Representation LearningHierarchical Metric LearningMulti-granularity FusionVision-Language ModelsPerson Retrieval
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Aerial-ground person re-identification must match the same person across cameras that look very different: close ground shots versus wide aerial shots. Earlier methods force features of the same identity from both views into one shared alignment, which erases cues that only one view actually provides. This paper argues that the right structure is a two-level hierarchy inside a hyperbolic space: a coarse parent level that separates identities and keeps them consistent across views, and a fine child level that is allowed to keep view-specific detail. Multi-granularity visual tokens from a vision-language encoder are first fused under dual text prompts, then mapped into that parent-child geometry and regularized by entailment cones so children stay inside their parents. On four public benchmarks the method reaches competitive or best mean average precision, showing that deliberately not collapsing view-specific information improves full-list ranking under severe viewpoint change.

Core claim

Direct cross-view alignment for aerial-ground person re-identification is suboptimal because it suppresses view-specific discriminative cues. By building hierarchical hyperbolic representations—coarse parent embeddings for identity separability and cross-view consistency, fine child embeddings for view-specific cues—regularized by Lorentz entailment constraints and supervised with mixed cross-view plus intra-view losses, the framework aggregates both kinds of features and improves retrieval.

What carries the argument

Hierarchical Hyperbolic Learning (HHL): view-agnostic fused features become scaled parent embeddings and view-aware fused features become farther child embeddings on a Lorentz manifold; four entailment-cone losses keep children inside parent cones and keep prompts and visuals consistent at each level.

Load-bearing premise

That casting view-agnostic features as hyperbolic parents and view-aware features as scaled children, then forcing children inside parent cones, is a faithful model of aerial-ground variation rather than an arbitrary regularizer.

What would settle it

On CARGO under the A→G protocol, train the identical multi-granularity backbone with and without the four entailment losses (or with parent and child roles swapped); if mAP does not fall when the cones are removed or inverted, the hierarchical geometry claim fails.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Preserving view-specific cues instead of erasing them can raise mean average precision on aerial-to-ground and ground-to-aerial protocols.
  • Hyperbolic space can host multi-granularity person features as a true hierarchy rather than a single flat embedding.
  • Text-guided fusion of intermediate Transformer layers supplies mid-level structure that global-only features miss.
  • Full-list ranking quality (mAP) improves when hierarchy prevents cross-view feature collapse.
  • The same coarse-to-fine hyperbolic pattern is offered as a template for other heterogeneous camera networks.

Where Pith is reading between the lines

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

  • Entailment cones may also suit other ReID settings where attributes form natural hierarchies such as clothing, pose, or time of day.
  • Because curvature and scale factors are learned end-to-end, similar hierarchies could be discovered for multi-modal retrieval without hand-assigned parent and child roles.
  • Aerial-ground gaps resemble other domain gaps (day-night, RGB-infrared); hierarchical hyperbolic alignment might reduce the need for synthetic view translation.
  • Practitioners could replace pure contrastive alignment losses with parent-child cone constraints whenever two views are systematically asymmetric.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 6 minor

Summary. The paper proposes HiHR for Aerial-Ground Person Re-Identification. It extracts multi-granularity visual features from CLIP Transformer layers together with dual-level view-agnostic and view-aware text prompts, fuses them via Text-guided Multi-granularity Fusion (TMF), and embeds the fused features into a Lorentz hyperbolic hierarchy via Hierarchical Hyperbolic Learning (HHL). Parent (view-agnostic) embeddings are supervised for cross-view identity consistency; child (view-aware) embeddings are supervised within-view and constrained inside parent entailment cones (Eqs. 7–13, §3.3). Inference concatenates the log-mapped parent and child features. Experiments on AG-ReID v1/v2, LAGPeR, and CARGO report competitive or best mAP under most protocols, with ablations of TMF, HHL, losses, prompts, fusion, layers, and hyperparameters.

Significance. AG-ReID is practically relevant because real surveillance mixes aerial and ground platforms with large viewpoint gaps. The paper’s main contribution is architectural: it is, to the authors’ knowledge, the first use of hierarchical hyperbolic embeddings for cross-view person retrieval, and it explicitly tries to keep view-specific cues rather than only aligning view-invariant features. Strengths include a clear modular design (MFE/TMF/HHL), consistent mAP gains across four public benchmarks, component and hyperparameter ablations (Tables 4–7, Fig. 4), and a public code link. If the parent–child entailment construction is a useful inductive bias, the work offers a reusable template for heterogeneous-view ReID beyond pure Euclidean alignment.

major comments (2)
  1. §3.3, Eqs. (7)–(13) and the four entailment terms in Eq. (16): the central claim that HHL “preserves view-specific discriminative cues” while enforcing identity hierarchy rests on treating view-agnostic features as parents and view-aware features as scaled children inside Lorentz cones. Ablations (Tables 4–5) show the losses help, but there is no direct diagnostic that the geometry retains view-specific information rather than acting as a generic regularizer (e.g., no cone-membership / aperture statistics, no parent–child distance analysis, no controlled comparison that freezes hierarchy while removing hyperbolic geometry or vice versa). A short analysis or alternative Euclidean hierarchical baseline would make the inductive-bias claim load-bearing rather than inferred from end metrics.
  2. Tables 1–3: the paper claims “competitive or state-of-the-art” performance, yet Rank-1 is not uniformly best (e.g., CARGO ALL Rank-1 74.36 vs LATex 76.96; several AG-ReID v1/v2 Rank-1 entries trail LATex/ViSA). Results are single-run with no multi-seed means or error bars, so the mAP leadership—especially the smaller margins—cannot be assessed for significance. Reporting mean±std over seeds (or at least confirming single-run protocol) is needed for the SOTA claim to be fully supported.
minor comments (6)
  1. Fig. 4 caption/labeling is duplicated in the manuscript text (“Fig. 4: Ablation study…” then again “Fig. 4: Hyperparameter and layer-selection…”); clean the figure numbering and captions.
  2. §3.1–3.2: notation for selected layers {l1…lM}, class tokens C, and fused tokens sk is dense; a short symbol table or expanded description of how αk is reused for patch aggregation would help reproducibility.
  3. §4.2: free parameters (λe, λg / λg,intra, s1, s2, τ, k, layer set) are listed with defaults; briefly state whether curvature and scales are optimized jointly with the backbone or with a different learning rate.
  4. Related work §2.2 correctly notes limited prior hyperbolic ReID; a one-sentence contrast with non-hierarchical Poincaré/kernel baselines (Khrulkov et al., Fang et al.) on AG-ReID would sharpen novelty.
  5. Minor prose issues: “manyAG-ReIDmethods”, “theidealalignment”, and similar spacing typos in early pages; also fix “https: //github.com” spacing in the abstract.
  6. Inference (§3.4) concatenates log-mapped parent and child features after undoing scales; state the final feature dimension and whether any re-normalization is applied before cosine/Euclidean matching.

Circularity Check

0 steps flagged

No significant circularity: HiHR is an empirical method paper whose hierarchy, losses, and reported mAP/Rank-1 gains are design choices evaluated on external public AG-ReID splits, not algebraic restatements of the metrics.

full rationale

The paper proposes Multi-granularity Feature Extraction, Text-guided Multi-granularity Fusion, and Hierarchical Hyperbolic Learning (parent/child scale separation, Lorentz exp/log maps, four entailment regularizers, and level-specific CE+triplet objectives). These are inductive design choices, not definitions of the evaluation metrics. Training objective L (Eq. 16) combines Lg(gp), Lg,intra(gc), and weighted entailment terms; inference concatenates log-mapped parent and child features. Reported Rank-1 and mAP are computed on held-out official splits of AG-ReID v1/v2, LAGPeR, and CARGO against external baselines. Ablations (Tables 4–7, Fig. 4) vary modules, losses, prompts, fusion, layers, and hyperparameters and show empirical gains; none of those quantities is forced by construction from the inputs. Self-citations (e.g., SAS-VPReID, LATex, SD-ReID) appear as related AG-ReID baselines or prior work, not as uniqueness theorems that forbid alternatives or as load-bearing premises that define the present results. Hyperbolic geometry and entailment cones are imported from the broader literature and used as a regularizer; they do not tautologically produce the claimed retrieval numbers. Therefore the derivation chain is self-contained against external benchmarks and exhibits no self-definitional, fitted-as-prediction, or self-citation-forced circularity.

Axiom & Free-Parameter Ledger

6 free parameters · 4 axioms · 3 invented entities

The central claim rests on standard hyperbolic geometry tools, the transferability of CLIP multi-layer features and dual prompts to AG-ReID, and an ad-hoc hierarchical assignment of view-agnostic vs view-aware features to parent/child cones. Several scalar hyperparameters control scale separation and loss balance; invented modules are architectural rather than physical entities.

free parameters (6)
  • entailment weight λ_e = 5
    Balances four entailment losses against ReID objectives in Eq. 16; default 5, chosen experimentally.
  • classification weights λ_g / λ_g,intra = 1.0 / 0.25
    Balance CE vs triplet at parent and child levels; set to 1.0 and 0.25 by default.
  • parent/child scale factors (s1, s2) = init (0.5, 2.0)
    Initialize hierarchical radial separation in Eq. 9; optimized end-to-end from (0.5, 2.0).
  • curvature τ = init 1.0
    Lorentz manifold curvature; initialized to 1.0 and optimized; sensitivity shown in Fig. 4.
  • cone aperture constant k = 0.1
    Fixed constant in half-aperture formula Eq. 7 controlling entailment cone width.
  • selected Transformer layers = {5,8,12}
    Multi-granularity extraction uses layers {5,8,12} chosen by ablation rather than derived.
axioms (4)
  • standard math Lorentz hyperbolic geometry with exponential/logarithmic maps and entailment cones is a valid embedding geometry for hierarchical visual features.
    Invoked throughout §3.3 via exp/log maps and cone aperture formulas from hyperbolic geometry literature.
  • domain assumption CLIP multi-layer visual tokens plus dual text prompts provide complementary multi-granularity identity cues for AG-ReID.
    Foundation of MFE/TMF in §3.1–3.2; relies on pretrained vision-language transfer without proving sufficiency for aerial-ground gaps.
  • ad hoc to paper View-agnostic features should be parents and view-aware features children in a coarse-to-fine hierarchy, with child embeddings constrained inside parent identity cones.
    Core modeling choice of HHL (§3.3, Fig. 3); not derived from data geometry, imposed by construction via scale separation and Le losses.
  • ad hoc to paper Cross-view mixed sampling at the parent level and same-view sampling at the child level correctly separate view-invariant vs view-specific supervision.
    Hierarchical supervision design in §3.3 (Lg vs Lg,intra); assumes this split avoids feature collapse without harming retrieval.
invented entities (3)
  • Text-guided Multi-granularity Fusion (TMF) no independent evidence
    purpose: Fuse multi-layer class/patch tokens under view-agnostic and view-aware text queries into parent/child Euclidean features.
    New module in §3.2; architectural construct validated only by ablations on CARGO, not by independent theory.
  • Hierarchical Hyperbolic Learning (HHL) parent-child structure no independent evidence
    purpose: Embed fused features as coarse identity parents and fine view-specific children with four entailment constraints.
    Central invented training structure in §3.3; evidence is internal benchmark gains, not external geometric measurement of person hierarchies.
  • Dual-level view-agnostic / view-aware prompts with [shared] and [view-private] tokens no independent evidence
    purpose: Provide semantic queries that separate identity-consistent and viewpoint-conditioned information.
    Prompt design in §3.1 and Table 6; dataset-level learnable tokens specific to this framework.

pith-pipeline@v1.1.0-grok45 · 17588 in / 3681 out tokens · 45381 ms · 2026-07-13T04:49:23.744565+00:00 · methodology

0 comments
read the original abstract

Aerial-Ground Person Re-IDentification (AG-ReID) aims to retrieve the same person across heterogeneous aerial and ground camera platforms. Although great progress has been made, existing methods remain suboptimal due to the direct feature alignment across views, overlooking view-specific cues. To address this issue, we propose a novel Hierarchical Hyperbolic Representation (HiHR) framework for AG-ReID. More specifically, we first extract multi-granularity features based on pre-trained visual-text encoders. Then, we propose a Text-guided Multi-granularity Fusion (TMF) to fuse multi-granularity features and enhance the representation ability of identity features. Furthermore, we introduce the Hierarchical Hyperbolic Learning (HHL) to construct a hierarchical feature structure in a hyperbolic space. This hierarchy includes a coarse level that ensures identity separability and cross-view consistency, and a fine level that preserves view-specific discriminative cues. As a result, our proposed framework can effectively aggregate view-invariant and view-specific discriminative features for AG-ReID. Extensive experiments on four AG-ReID benchmarks demonstrate the effectiveness of our framework. The source code is available at https://github.com/YangQiWei3/HiHR.

Figures

Figures reproduced from arXiv: 2607.09186 by Pingping Zhang, Qiwei Yang.

Figure 1
Figure 1. Figure 1: Illustration of our motivations and proposed framework. (a) The ideal alignment clusters examples of the same identity nearby while different identities far apart. (b) Previous alignments in Euclidean space align cross-view samples directly and may suppress view-specific discriminative cues. (c) Our hierarchical hyperbolic alignment keeps a hierarchical separability and view-specific discriminative cues. T… view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of our proposed framework. It includes three key modules, i.e., Multi-granularity Feature Extraction (MFE), Text-guided Multi-granularity Fusion (TMF) and Hierarchical Hyperbolic Learning (HHL). explored. Early attempts mainly adopt hyperbolic geometry as a drop-in re￾placement of the Euclidean embedding space. For example, Khrulkov et al. [8] append hyperbolic layers to standard visual encode… view at source ↗
Figure 3
Figure 3. Figure 3: Illustration of entailment regularizations for hierarchical consistency in HHL. Four entailment constraints are imposed: parent prototypes to child embeddings (z π → z c ), parent prompts to child prompts (z p P → z c P ), parent prompts to parent visual features (z p P →z p ), and child prompts to child visual features (z c P →z c ). These constraints jointly preserve hierarchical consistency, align visio… view at source ↗
Figure 4
Figure 4. Figure 4: Ablation study on the hyperparameters on CARGO. Fig. 4: Hyperparameter and layer-selection analysis on CARGO. [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗

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

35 extracted references · 3 linked inside Pith

  1. [1]

    Flavors of geometry31(59-115), 2 (1997)

    Cannon, J.W., Floyd, W.J., Kenyon, R., Parry, W.R., et al.: Hyperbolic geometry. Flavors of geometry31(59-115), 2 (1997)

    1997

  2. [2]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Fang, P., Harandi, M., Petersson, L.: Kernel methods in hyperbolic spaces. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 10665–10674 (2021)

    2021

  3. [3]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Ha, R., Jiang, S., Li, B., Pan, B., Zhu, Y., Zhang, J., Zhu, X., Gong, S., Wang, J.: Multi-modal multi-platform person re-identification: Benchmark and method. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 10251–10261 (2025)

    2025

  4. [4]

    arXiv preprint arXiv:1703.07737 (2017)

    Hermans, A., Beyer, L., Leibe, B.: In defense of the triplet loss for person re- identification. arXiv preprint arXiv:1703.07737 (2017)

    Pith/arXiv arXiv 2017

  5. [5]

    In: International Conference on Image and Graphics

    Huang, Y., Chen, H., Feng, Z., Lai, J.: Perspective driven prototype alignment for aerial-ground person re-identification. In: International Conference on Image and Graphics. pp. 517–528. Springer (2025)

    2025

  6. [6]

    In: Proceedings of the IEEE/CVF International Conference on Com- puter Vision

    Huang, Z., Liu, X.: Generalizable object re-identification via visual in-context prompting. In: Proceedings of the IEEE/CVF International Conference on Com- puter Vision. pp. 22539–22550 (2025)

    2025

  7. [7]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Khalid, W., Liu, B., Li, X., Waqas, M., Afgan, M.S.: Bridging the sky and ground: Towards view-invariant feature learning for aerial-ground person re-identification. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 9749–9758 (2025)

    2025

  8. [8]

    In: Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition

    Khrulkov, V., Mirvakhabova, L., Ustinova, E., Oseledets, I., Lempitsky, V.: Hyper- bolic image embeddings. In: Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition. pp. 6418–6428 (2020)

    2020

  9. [9]

    arXiv preprint arXiv:1412.6980 (2014)

    Kingma, D.P., Ba, J.: Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014)

    Pith/arXiv arXiv 2014

  10. [10]

    Springer Science & Business Media (2006)

    Lee, J.M.: Riemannian manifolds: an introduction to curvature. Springer Science & Business Media (2006)

    2006

  11. [11]

    Advances in Neural Information Processing Systems38, 93785–93805 (2026)

    Li, Q., Li, J., Zhang, Y., Tan, L., Chen, J., Ji, J.: Gsalign: Geometric and semantic alignment network for aerial-ground person re-identification. Advances in Neural Information Processing Systems38, 93785–93805 (2026)

    2026

  12. [12]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Li, S., Sun, L., Li, Q.: Clip-reid: exploiting vision-language model for image re- identification without concrete text labels. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 37, pp. 1405–1413 (2023)

    2023

  13. [13]

    ACM Transactions on Multimedia Computing, Communications and Applications22(3), 1–19 (2026)

    Mao, D., Teng, S., Lyu, X.: Cvaf: A clip-based view-consistent alignment frame- work for aerial-ground person re-identification. ACM Transactions on Multimedia Computing, Communications and Applications22(3), 1–19 (2026)

    2026

  14. [14]

    In: IEEE International Conference on Multimedia and Expo

    Nguyen, H., Nguyen, K., Sridharan, S., Fookes, C.: Aerial-ground person re-id. In: IEEE International Conference on Multimedia and Expo. pp. 2585–2590 (2023)

    2023

  15. [15]

    v2: Bridging aerial and ground views for person re-identification

    Nguyen, H., Nguyen, K., Sridharan, S., Fookes, C.: Ag-reid. v2: Bridging aerial and ground views for person re-identification. IEEE Transactions on Information Forensics and Security19, 2896–2908 (2024)

    2024

  16. [16]

    In: IEEE International Joint Conference on Bio- metrics

    Nguyen, K., Fookes, C., Sridharan, S., Liu, F., Liu, X., Ross, A., Michalski, D., Nguyen, H., Deb, D., Kothari, M., et al.: Ag-reid 2023: Aerial-ground person re- identification challenge results. In: IEEE International Joint Conference on Bio- metrics. pp. 1–10 (2023)

    2023

  17. [17]

    Yang models

    Pal, A., Van Spengler, M., D’Amely di Melendugno, G., Flaborea, A., Galasso, F., Mettes, P.: Compositional entailment learning for hyperbolic vision-language 16 Q. Yang models. In: International Conference on Learning Representations. vol. 2025, pp. 87371–87399 (2025)

    2025

  18. [18]

    In: International Conference on Machine Learning

    Radford, A., Kim, J.W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., et al.: Learning transferable visual models from natural language supervision. In: International Conference on Machine Learning. pp. 8748–8763 (2021)

    2021

  19. [19]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

    Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., Wojna, Z.: Rethinking the incep- tion architecture for computer vision. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. pp. 2818–2826 (2016)

    2016

  20. [20]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Wang, S., Wang, Y., Wu, R., Jiao, B., Wang, W., Wang, P.: Secap: self-calibrating and adaptive prompts for cross-view person re-identification in aerial-ground net- works. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 22119–22128 (2025)

    2025

  21. [21]

    IEEETransactions on Circuitsand Systems for Video Technology pp

    Wang, Y., Zhang, P., Sun, C., Wang, D., Lu, H.: What makes you unique? attribute prompt composition forobject re-identification. IEEETransactions on Circuitsand Systems for Video Technology pp. 3173–3184 (2025)

    2025

  22. [22]

    Computer Vision and Image Understanding244, 104030 (2024)

    Wang, Y., Zhang, P., Wang, D., Lu, H.: Other tokens matter: Exploring global and local features of vision transformers for object re-identification. Computer Vision and Image Understanding244, 104030 (2024)

    2024

  23. [23]

    arXiv preprint arXiv:2412.00433 (2024)

    Wang, Y., Pishgar, M.: Dynamic token selection for aerial-ground person re- identification. arXiv preprint arXiv:2412.00433 (2024)

    Pith/arXiv arXiv 2024

  24. [24]

    IEEE Transactions on Image Processing 35, 5686–5697 (2026)

    Wang, Y., Hu, X., Wang, L., Zhang, P., Lu, H.: Sd-reid: View-aware stable diffusion for aerial-ground person re-identification. IEEE Transactions on Image Processing 35, 5686–5697 (2026)

    2026

  25. [25]

    In: Proceedings of the AAAI Con- ference on Artificial Intelligence

    Wang, Y., Liu, Y., Zheng, A., Zhang, P.: Decoupled feature-based mixture of ex- perts for multi-modal object re-identification. In: Proceedings of the AAAI Con- ference on Artificial Intelligence. vol. 39, pp. 8141–8149 (2025)

    2025

  26. [26]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Wang, Y., Lv, Y., Zhang, P., Lu, H.: Idea: Inverted text with cooperative de- formable aggregation for multi-modal object re-identification. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 29701–29710 (2025)

    2025

  27. [27]

    IEEE Trans- actions on Intelligent Transportation Systems pp

    Wang, Y., Zhang, P., Liu, X., Tu, Z., Lu, H.: Unity is strength: Unifying convolu- tional and transformeral features for better person re-identification. IEEE Trans- actions on Intelligent Transportation Systems pp. 3713–3723 (2025)

    2025

  28. [28]

    Visual Intelligence1(1), 24 (2023)

    Yan, P., Liu, X., Zhang, P., Lu, H.: Learning convolutional multi-level transformers for image-based person re-identification. Visual Intelligence1(1), 24 (2023)

    2023

  29. [29]

    In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision Workshops

    Yang, Q., Zhang, P., Wang, Y., Gong, Z.: Sas-vpreid: A scale-adaptive framework with shape priors for video-based person re-identification at extreme far distances. In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision Workshops. pp. 1599–1608 (March 2026)

    2026

  30. [30]

    arXiv preprint arXiv:2503.23722 (2025)

    Zhang, P., Hu, X., Wang, Y., Lu, H.: Latex: Leveraging attribute-based text knowl- edge for aerial-ground person re-identification. arXiv preprint arXiv:2503.23722 (2025)

    arXiv 2025

  31. [31]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Zhang, P., Wang, Y., Liu, Y., Tu, Z., Lu, H.: Magic tokens: Select diverse tokens for multi-modal object re-identification. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 17117–17126 (2024)

    2024

  32. [32]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Zhang, Q., Cai, Z., Zhao, P., Wu, J., Wu, C., Chen, H., Lai, J.: View-aware se- mantic alignment for aerial-ground person re-identification. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 4383– 4392 (2026) HiHR: Hierarchical Hyperbolic Representation for AG-ReID 17

    2026

  33. [33]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Zhang, Q., Wang, L., Patel, V.M., Xie, X., Lai, J.: View-decoupled transformer for person re-identification under aerial-ground camera network. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 22000–22009 (2024)

    2024

  34. [34]

    In: Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition

    Zhou, K., Yang, J., Loy, C.C., Liu, Z.: Conditional prompt learning for vision- language models. In: Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition. pp. 16816–16825 (2022)

    2022

  35. [35]

    In: Chinese Conference on Pattern Recognition and Computer Vision

    Zhu, R., Chen, H., Xie, X., Lai, J.: Global-local prompts-driven semantic guid- ance for aerial-ground person re-identification. In: Chinese Conference on Pattern Recognition and Computer Vision. pp. 99–112. Springer (2025)

    2025