Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

ICPL-ReID: Identity-Conditional Prompt Learning for Multi-Spectral Object Re-Identification

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Identity text prompts, trained online inside CLIP, outperform prior multi-spectral re-identification methods on five benchmarks.

desk verdict A solid multi-spectral ReID recipe with consistent gains, but the CLIP-semantics story needs a control and the evaluation needs variance reporting. read the letter →

arxiv 2505.17821 v1 pith:PCWPMDTN submitted 2025-05-23 cs.CV

Shihao Li , Chenglong Li , Aihua Zheng , Jin Tang , Bin Luo This is my paper · ORCID
classification cs.CV
keywords multi-spectralre-identificationpromptlearningCLIPidentity-conditionedpromptslow-rankadapterRGB-NIR-TIRpersonvehicle
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper claims that CLIP's pre-trained image-text alignment space can act as a universal bridge for multi-spectral re-identification, so that RGB, near-infrared, and thermal features of the same identity become comparable without designing intricate fusion modules. It proposes ICPL, which learns per-identity text prompts for each spectrum and trains them online together with a lightly adapted CLIP visual encoder. To keep this online learning stable, the prompts are conditioned on momentum-updated identity prototypes, and a low-rank adapter is inserted into each transformer block to fit small-scale spectral data without disturbing the pre-trained alignment. Experiments on RGBNT201, Market-MM, MSVR310, RGBN300, and RGBNT100 report higher mAP and Rank-1 than prior state-of-the-art methods, including the two-stage CLIP-ReID. If this holds, a frozen vision-language model can be repurposed for heterogeneous spectral identity matching without real text labels.

What carries the argument

The load-bearing mechanism is the identity-conditioned alignment loop. It combines image-to-text contrastive loss $\mathcal{L}_{i2t}$, text-to-prototype and prototype-to-text losses $\mathcal{L}_{t2p}$ and $\mathcal{L}_{p2t}$ computed against momentum-updated identity prototypes, an image-to-prototype loss $\mathcal{L}_{i2p}$, and identity and triplet losses. The multi-spectral adapter is a residual low-rank bottleneck added to the feed-forward network of each ViT block with $W_{down}$, ReLU, $W_{up}$, and a scaling factor $s$, allowing the frozen CLIP encoder to absorb spectral style shifts with roughly 11-17% of full fine-tuning parameters.

What would settle it

Run the same ICPL training but replace the text branch with random, non-textual learned vectors of the same dimension per identity. If mAP and Rank-1 on RGBNT201 and MSVR310 stay at roughly 75.1/77.4 and 56.9/77.7, the prompt losses are not carrying semantic information. Conversely, permuting the assignment of prompts to identities should destroy performance if the semantics matter.

Watch

Extended reading notes

Core claim

The central claim is that identity-level text semantics can unify heterogeneous spectral views: for each identity, ICPL builds a learnable prompt per spectrum and aligns it with spectral features through contrastive losses, replacing raw instance features with identity prototypes in the prompt-alignment loop. The paper argues that because the prototypes condense each identity into a many-to-one anchor, the online mutual optimization of text prompt and visual encoder avoids the semantic shift seen in two-stage pre-alignment methods. A low-rank adapter in each visual transformer block freezes most of CLIP while learning spectra-specific features. At inference the text branch is discarded and only the concatenated spectral features are used, so the entire prompt mechanism is a training-time regularizer that the paper reports as outperforming all compared methods on five benchmarks.

Load-bearing premise

The load-bearing premise is that CLIP's pre-trained image-text embedding space, learned on natural RGB images, still gives a meaningful geometry for aligning near-infrared and thermal identity features to learnable text prompts, even though no real text-image pairs for those identities exist.

Editorial extensions

If this is right

  • If the claim is right, multi-spectral ReID no longer needs dedicated cross-modal fusion modules; a frozen CLIP encoder plus small adapters and training-time prompts suffices.
  • The two-stage prompt paradigm, where prompts are pre-aligned and then the visual encoder is fine-tuned separately, is suboptimal; end-to-end identity-conditioned prompting is the stronger recipe.
  • The same identity-conditioned prompt recipe should transfer to other spectra, such as depth or event data, where no text labels exist.
  • Low-rank adapters preserve the pre-trained alignment distribution while using a fraction of tunable parameters, making the approach practical on small multi-spectral datasets.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A testable consequence the paper leaves implicit: if the learned prompts are genuinely semantic, they should support zero-shot or few-shot retrieval for unseen identities by reusing the text encoder with new prompt tokens; this is not evaluated.
  • The gains could partly come from the extra prototype-contrastive metric losses rather than from CLIP's text semantics; an ablation replacing the text branch with random learned vectors of equal dimension would settle how much of the improvement is language-driven.
  • The identity-prototype many-to-one trick is a general regularizer for online prompt learning and could be applied to other label-scarce multimodal retrieval tasks, such as unsupervised ReID with pseudo-labels.
  • Because Market-MM is synthetic, its reported margin may not reflect real sensor noise; a real-world NIR/TIR deployment test would be needed to confirm the gains.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes ICPL-ReID, a CLIP-based framework for multi-spectral person and vehicle re-identification. It freezes a CLIP text encoder and most of a CLIP visual encoder, learns identity-specific text prompts online, aggregates per-identity spectral prototypes with momentum, aligns prompts to prototypes and instances through several contrastive losses, and inserts low-rank adapters to learn spectral-specific features. At inference only concatenated spectral visual features are used. Experiments on RGBNT201, Market-MM, MSVR310, RGBN300, and RGBNT100 report improvements over prior methods including CLIP-ReID, with component ablations and hyperparameter analyses.

Significance. If validated, the paper would show that parameter-efficient prompt learning can transfer CLIP-style cross-modal alignment to multi-spectral ReID, which is practically useful given the small scale and stylistic diversity of multi-spectral datasets. The benchmarking is broad (five datasets, person and vehicle) and the ablations are systematic; public code and a computational-efficiency comparison are also strengths. However, the central attribution to CLIP semantics is not yet established, because the learnable prompts are optimized against identity prototypes derived from the same visual features without any external text grounding, and the statistical evidence is weakened by single-run results and test-set hyperparameter scanning. The zero mAP gain on RGBNT100 also conflicts with the abstract's unqualified SOTA claim.

major comments (4)
  1. [Sec. III.B-C, Eqs. (1), (6)-(7)] The central mechanistic claim that ICPL 'exploits the powerful cross-modal alignment capability of CLIP' (Abstract) is not supported by the experiments as reported. The learnable identity prompts t_c_m are optimized with L_i2t and L_t2p/L_p2t against identity prototypes u_c_m that are themselves running averages of the same visual features (Eqs. (3)-(4)). Because no real text-image pairs are available for these identities, and Section III.C concedes that 'the model cannot observe the real spectral and text alignment distribution,' the losses can act as label-conditioned contrastive regularizers whose positive/negative structure is determined entirely by identity labels. The paper needs a control experiment that replaces the frozen pre-trained CLIP text encoder with a randomly initialized text encoder (or removes it while preserving the same prototype/instance contrastive losses and the adapter). If the gains persist, the improvement should be attributed to the metric-learning structure rather than to CLIP semantics; if they disappear, the paper should report that. Without such a control, the attribution in the title and abstract is not established.
  2. [Table IV and Sec. IV.C (RGBNT100)] The abstract and Section IV.C claim state-of-the-art results on all five benchmarks, but Table IV reports an mAP of 87.0 for both ICPL and CLIP-ReID on RGBNT100. The zero mAP gain is acknowledged and explained by repeated viewpoints, but the claim 'outperforms the state-of-the-art' is overstated for that benchmark, and the explanation is post hoc. Please report the exact margin, provide variance over repeated runs, and either soften the claim or demonstrate with additional metrics that the Rank-1 gain is stable. This is load-bearing because the paper's headline contribution is uniform SOTA improvement.
  3. [Sec. IV.B-IV.E, Tables I-IV, VII-IX] All experimental tables report a single run with no standard deviation or seed information, and many hyperparameters (alpha in Fig. 6, adapter scale s in Table VII, prompt number M in Table VIII, and loss weights lambda_1, lambda_2, lambda_3 in Table IX) are selected by scanning final test-set metrics. On small datasets where the reported differences are often 1-2 mAP, this makes it difficult to distinguish genuine improvements from selection noise. Please report mean and standard deviation over at least three seeds, or clearly state that the differences are within noise. This is particularly important for the RGBNT100 tie and for the small gains in Tables V and VIII.
  4. [Eqs. (3)-(4) and Algorithm 1] The prototype update mechanism is not fully specified. Eq. (4) shows a momentum update with a single instance feature v^{c,j}_m, and Algorithm 1 line 14 updates U with v^i, but the text preceding Eq. (3) states that prototypes are aggregated 'before each training epoch.' It is unclear whether the epoch-beginning aggregate is recomputed from the full training set with the current encoder, whether the memory bank is updated only with the sampled batch, and whether the prototype for a batch is the same for all samples of an identity. Please state the exact schedule and update rule; this matters because the prompt-text alignment losses (Eqs. (6)-(7)) are computed against these prototypes and the method's stability depends on them.
minor comments (6)
  1. [Notation, Eqs. (1)-(7) and Table VIII] The symbol M is used both for the number of spectra in Eqs. (1)-(7) and for the number of learnable prompt tokens in Table VIII; please use distinct symbols (e.g., M_s and M_t).
  2. [Sec. III.B and Sec. IV.E] In Section III.B, 'triplet-steam visual encoder' should read 'triplet-stream visual encoder,' and in Section IV.E 'RGBNT210' should read 'RGBNT201.'
  3. [Table III and references] The AGW method in Table III cites reference [66], which is the TPAMI survey by Ye et al. rather than the AGW paper; please correct the citation.
  4. [Fig. 5] The x-axis is labeled 'Number of tunable parameters (M)' but the values (0.5 to 512) appear to be the adapter bottleneck dimension d-tilde, not millions of parameters; please relabel to avoid confusion.
  5. [Sec. IV.D] The phrase 'cross three datasets' appears several times and should be 'across three datasets'; there are also minor spacing issues such as 'L_{final}' in Eq. (12).
  6. [Algorithm 1] Line 14 (T^{i+1}=t^i) does not match the description in the text that the text prompt tokens [X]_m are updated via the text encoder; please clarify whether T^i denotes the token embeddings or the encoded text features.

Circularity Check

1 steps flagged · score 2.0 of 10

The learnable 'text semantics' are self-derived from the same visual features they are claimed to teach, but the SOTA evaluation is external and prompt-free at inference, so the circularity is limited.

  1. self definitional [Section III.C, Eqs. (3), (6), (7), and (8)]
    "Due to the lack of concrete text descriptions, the model cannot observe the real spectral and text alignment distribution... before each training epoch, we aggregate image features belonging to the same identity to yield multi-spectral prototypes U_m... u_c^m = 1/N_c^m \sum_j v^{c,j}_m ... employ the contrastive loss L_t2p and L_p2t to align the text prompt with the prototypes."

    The text prompt t_c^m is presented as an identity-level semantic center that bridges identity semantics across spectra, but its only supervision in L_t2p and L_p2t is the prototype u_c^m, which Eq. (3) defines as the mean of the same spectral visual features v^{c,j}_m. Thus the 'text semantics' are the model's own visual feature averages re-encoded through the frozen CLIP text transformer; the prompt carries no independent linguistic or externally grounded identity information. The alignment loop is closed over the model's own outputs, so the claim that text semantics guide spectral alignment is self-derived by construction.

full rationale

The paper's central empirical claim is an external benchmark comparison, and the claimed SOTA gains are measured on held-out test splits of five public datasets; no test labels, gallery rankings, or evaluation metrics are fitted through the prompt loop, and the text prompt is discarded at inference, with only concatenated spectral features used for retrieval. The only identifiable circularity is in the mechanism narrative: the learnable text prompt is optimized to match identity prototypes that are themselves arithmetic means of the same spectral features (Eq. 3 with Eqs. 6-7), so the 'text semantics' are self-referential rather than grounded in real language. However, this self-referential training signal is a regularizer during training, not a claimed prediction; the absence of a random-text-encoder control makes the CLIP-semantics attribution uncertain, but that is a correctness/ablation concern, not a derivation that makes the benchmark outcome circular. There is no load-bearing self-citation chain and no fitted parameter renamed as a prediction. Accordingly, the circularity score is low: one self-definitional component in the stated mechanism, while the central evaluation remains externally grounded.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

The method depends on several domain assumptions about CLIP transfer and prototype stability, plus seven tuned hyperparameters. No fundamentally new physical or mathematical entity is introduced; the identity prototypes are running averages of training features rather than independent inventions.

free parameters (7)
  • learnable identity text prompts t_c_m = learned per identity per spectrum; M=4 prompt tokens selected from Table VIII
    Core online prompt learning parameters; they are optimized against identity prototypes derived from the same visual features (Eqs. 6-8), so they carry no external text supervision.
  • contrastive temperature gamma = not reported
    Used in all contrastive alignment losses (Eqs. 1, 2, 5-7); value is a hyperparameter chosen by the authors but never stated in the text.
  • loss weights lambda1, lambda2, lambda3 = lambda1=0.1, lambda2=1.0, lambda3=0.9 (default settings in Table IX)
    Balancing weights for L_i2t, prototype-text alignment, and image-prototype loss; selected from test-set sweeps in Table IX.
  • prototype momentum factor alpha = not stated; swept in Fig. 6
    Controls prototype update speed (Eq. 4); the paper gives no explicit default value in the implementation section.
  • adapter scaling factor s = per-dataset from Table VII (e.g., 0.5 on RGBNT201, 0.3-0.4 on MSVR310)
    Controls the strength of the adapter residual (Eq. 9); selected by sweeping values from 0.1 to 1.0.
  • adapter bottleneck dimension d_tilde = d_tilde=512 (shared adapter, MSVR310), d_tilde=768 (per-modality, RGBNT201)
    Intermediate adapter dimension in Fig. 5; chosen from {16,...,768} to balance parameters and accuracy.
  • number of prompt tokens M = M=4
    Selected from sweep of {1,2,4,8,16,32} in Table VIII.
assumptions (4)
  • domain assumption CLIP's pre-trained RGB image-text embedding space remains a valid alignment space for NIR and TIR identity features after low-rank adaptation.
    Invoked throughout Section III.B-C; the method relies on aligning spectral features to learnable text prompts without any concrete text-image pairs.
  • domain assumption Identity prototypes formed by averaging instance features (Eq. 3) and momentum updates (Eq. 4) are stable, identity-discriminative anchors for prompt learning.
    Section III.C uses these prototypes as the condition for prompt alignment; their validity is asserted, not derived.
  • domain assumption A lightweight low-rank adapter in each transformer FFN can adapt to a new spectral modality without destroying CLIP's pre-trained alignment distribution.
    Section III.D and Fig. 4 assume the adapter residual (Eq. 9) preserves the frozen backbone's generalization while adding spectral-specific features.
  • domain assumption Standard ReID losses (identity classification and triplet) remain effective when combined with contrastive text-prompt losses in a single end-to-end optimization.
    Section III.E defines the final loss as a sum of L_id, L_tri, L_i2p and L_prompt without proof of optimization stability.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ICPL-ReID: Identity-Conditional Prompt Learning for Multi-Spectral Object Re-Identification." pith.science (2026). https://pith.science/paper/PCWPMDTN

@misc{pith2026250517821,
  author       = {Pith},
  title        = {Pith review of: ICPL-ReID: Identity-Conditional Prompt Learning for Multi-Spectral Object Re-Identification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PCWPMDTN}},
  note         = {Machine review of arXiv:2505.17821}
}
read the original abstract

Multi-spectral object re-identification (ReID) brings a new perception perspective for smart city and intelligent transportation applications, effectively addressing challenges from complex illumination and adverse weather. However, complex modal differences between heterogeneous spectra pose challenges to efficiently utilizing complementary and discrepancy of spectra information. Most existing methods fuse spectral data through intricate modal interaction modules, lacking fine-grained semantic understanding of spectral information (\textit{e.g.}, text descriptions, part masks, and object keypoints). To solve this challenge, we propose a novel Identity-Conditional text Prompt Learning framework (ICPL), which exploits the powerful cross-modal alignment capability of CLIP, to unify different spectral visual features from text semantics. Specifically, we first propose the online prompt learning using learnable text prompt as the identity-level semantic center to bridge the identity semantics of different spectra in online manner. Then, in lack of concrete text descriptions, we propose the multi-spectral identity-condition module to use identity prototype as spectral identity condition to constraint prompt learning. Meanwhile, we construct the alignment loop mutually optimizing the learnable text prompt and spectral visual encoder to avoid online prompt learning disrupting the pre-trained text-image alignment distribution. In addition, to adapt to small-scale multi-spectral data and mitigate style differences between spectra, we propose multi-spectral adapter that employs a low-rank adaption method to learn spectra-specific features. Comprehensive experiments on 5 benchmarks, including RGBNT201, Market-MM, MSVR310, RGBN300, and RGBNT100, demonstrate that the proposed method outperforms the state-of-the-art methods.

Figures

Figures reproduced from arXiv: 2505.17821 by the authors.

Figure 1
Figure 1. (a) Classical pre-training models require ReID learning with id [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Pipeline of our proposed framework. (a) For end-to-end training of multi-spectral ReID, online prompt learning leverages learnable text prompt as [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Illustration of traditional instance feature learning strategy. (a) The [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Architecture of our multi-spectral adapter. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: The performance trend on mAP and Rank-1 as the number of tunable [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 7
Figure 7. Figure 7: Visualization results of the (a) Baseline, (b) Baseline + MS-A, (c) [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]
Figure 8
Figure 8. Figure 8: Visualization results of the (a) Baseline, (b) Baseline + MS-A, (c) Baseline + SIC + AL, and (d) Baseline + SIC + AL + MS-A (Ours), drawn by [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: Visualization of top-5 failure cases on (a) RGBNT201 [11], (b) MSVR310 [19], and (c) RGBNT100 [10] datasets. [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Multi-Modal Object Re-Identification with Dual Semantic Guidance and Global-Local Mutual Modulation

    cs.CV 2026-07 conditional novelty 4.0 of 10

    A dual-semantic (text + soft mask) global-local mutual modulation framework reports SOTA mAP/Rank-1 on RGBNT201, RGBNT100, and MSVR310.

Reference graph

Works this paper leans on

75 extracted references · 67 canonical work pages · cited by 1 Pith paper

  1. [1]

    A survey of open-world person re- identification,

    Q. Leng, M. Ye, and Q. Tian, “A survey of open-world person re- identification,”IEEE Transactions on Circuits and Systems for Video Technology, vol. 30, no. 4, pp. 1092–1108, 2020

    work page 2020

  2. [3]

    Tran- sreid: Transformer-based object re-identification,

    S. He, H. Luo, P. Wang, F. Wang, H. Li, and W. Jiang, “Tran- sreid: Transformer-based object re-identification,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 14 993–15 002

    work page 2021

  3. [4]

    Scalable person re-identification: A benchmark,

    L. Zheng, L. Shen, L. Tian, S. Wang, J. Wang, and Q. Tian, “Scalable person re-identification: A benchmark,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2015, pp. 1116–1124

    2015

  4. [5]

    Bag of tricks and a strong baseline for deep person re-identification,

    H. Luo, Y . Gu, X. Liao, S. Lai, and W. Jiang, “Bag of tricks and a strong baseline for deep person re-identification,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2019, pp. 1487–1495

    work page 2019

  5. [6]

    Omni-scale feature learning for person re-identification,

    K. Zhou, Y . Yang, A. Cavallaro, and T. Xiang, “Omni-scale feature learning for person re-identification,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2019, pp. 3702–3712

    work page 2019

  6. [7]

    Searching parameterized retrieval & verification loss for re-identification,

    X. Yuan, X. Xu, Z. Wang, K. Zhang, W. Liu, and R. Hu, “Searching parameterized retrieval & verification loss for re-identification,”IEEE Journal of Selected Topics in Signal Processing, vol. 17, no. 3, pp. 560–574, 2023

    work page 2023

  7. [8]

    Duality-gated mutual condition network for RGBT tracking,

    A. Lu, C. Qian, C. Li, J. Tang, and L. Wang, “Duality-gated mutual condition network for RGBT tracking,”IEEE Transactions on Neural Networks and Learning Systems, vol. 36, no. 3, pp. 4118–4131, 2025

    work page 2025

  8. [9]

    RGBT tracking via multi- adapter network with hierarchical divergence loss,

    A. Lu, C. Li, Y . Yan, J. Tang, and B. Luo, “RGBT tracking via multi- adapter network with hierarchical divergence loss,”IEEE Transactions on Image Processing, vol. 30, pp. 5613–5625, 2021

    work page 2021

Show all 75 references
  1. [10]

    Multi-spectral vehicle re-identification: A challenge,

    H. Li, C. Li, X. Zhu, A. Zheng, and B. Luo, “Multi-spectral vehicle re-identification: A challenge,” inProceedings of the AAAI Conference on Artificial Intelligence, 2020, pp. 11 345–11 353

    2020

  2. [11]

    Robust multi-modality person re-identification,

    A. Zheng, Z. Wang, Z. Chen, C. Li, and J. Tang, “Robust multi-modality person re-identification,” inProceedings of the AAAI Conference on Artificial Intelligence, 2021, pp. 3529–3537

    2021

  3. [12]

    Grayscale enhancement colorization network for visible-infrared person re- identification,

    X. Zhong, T. Lu, W. Huang, M. Ye, X. Jia, and C. Lin, “Grayscale enhancement colorization network for visible-infrared person re- identification,”IEEE Transactions on Circuits and Systems for Video Technology, vol. 32, no. 3, pp. 1418–1430, 2022

    2022

  4. [13]

    Modality-missing rgbt tracking: Invertible prompt learning and high-quality benchmarks,

    A. Lu, C. Li, J. Zhao, J. Tang, and B. Luo, “Modality-missing rgbt tracking: Invertible prompt learning and high-quality benchmarks,”In- ternational Journal of Computer Vision, pp. 1–21, 2024

    2024

  5. [14]

    Clip-reid: Exploiting vision-language model for image re-identification without concrete text labels,

    S. Li, L. Sun, and Q. Li, “Clip-reid: Exploiting vision-language model for image re-identification without concrete text labels,” inProceedings of the AAAI Conference on Artificial Intelligence, 2023, pp. 1405–1413

    2023

  6. [15]

    Learning discriminative features with multiple granularities for person re-identification,

    G. Wang, Y . Yuan, X. Chen, J. Li, and X. Zhou, “Learning discriminative features with multiple granularities for person re-identification,” in Proceedings of the ACM International Conference on Multimedia, 2018, pp. 274–282

    2018

  7. [16]

    Beyond part models: Person retrieval with refined part pooling (and A strong convolutional baseline),

    Y . Sun, L. Zheng, Y . Yang, Q. Tian, and S. Wang, “Beyond part models: Person retrieval with refined part pooling (and A strong convolutional baseline),” inProceedings of the European Conference on Computer Vision, vol. 11208, 2018, pp. 501–518

    2018

  8. [17]

    Il- lumination distillation framework for nighttime person re-identification and a new benchmark,

    A. Lu, Z. Zhang, Y . Huang, Y . Zhang, C. Li, J. Tang, and L. Wang, “Il- lumination distillation framework for nighttime person re-identification and a new benchmark,”IEEE Transactions on Multimedia, vol. 26, pp. 406–419, 2024

    2024

  9. [18]

    Sjdl- vehicle: Semi-supervised joint defogging learning for foggy vehicle re-identification,

    W. Chen, I. Chen, C. Yeh, H. Yang, J. Ding, and S. Kuo, “Sjdl- vehicle: Semi-supervised joint defogging learning for foggy vehicle re-identification,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 36, no. 1, 2022, pp. 347–355

    2022

  10. [19]

    Cross-directional consistency network with adaptive layer normalization for multi-spectral vehicle re-identification and a high-quality benchmark,

    A. Zheng, X. Zhu, Z. Ma, C. Li, J. Tang, and J. Ma, “Cross-directional consistency network with adaptive layer normalization for multi-spectral vehicle re-identification and a high-quality benchmark,”Information Fusion, vol. 100, p. 101901, 2023

    2023

  11. [20]

    Heterogeneous test-time training for multi-modal person re-identification,

    Z. Wang, H. Huang, A. Zheng, and R. He, “Heterogeneous test-time training for multi-modal person re-identification,” inProceedings of the AAAI Conference on Artificial Intelligence, 2024, pp. 5850–5858

    2024

  12. [21]

    Top-reid: Multi- spectral object re-identification with token permutation,

    Y . Wang, X. Liu, P. Zhang, H. Lu, Z. Tu, and H. Lu, “Top-reid: Multi- spectral object re-identification with token permutation,” inProceedings of the AAAI Conference on Artificial Intelligence, 2024, pp. 5758–5766

    2024

  13. [22]

    Magic tokens: Select di- verse tokens for multi-modal object re-identification,

    P. Zhang, Y . Wang, Y . Liu, Z. Tu, and H. Lu, “Magic tokens: Select di- verse tokens for multi-modal object re-identification,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 17 117–17 126

    2024

  14. [23]

    Empowering visible-infrared person re- identification with large foundation models,

    Z. Hu, B. Yang, and M. Ye, “Empowering visible-infrared person re- identification with large foundation models,” inThe Thirty-eighth Annual Conference on Neural Information Processing Systems, 2024

    2024

  15. [24]

    Multi- prompts learning with cross-modal alignment for attribute-based person re-identification,

    Y . Zhai, Y . Zeng, Z. Huang, Z. Qin, X. Jin, and D. Cao, “Multi- prompts learning with cross-modal alignment for attribute-based person re-identification,” inProceedings of the AAAI Conference on Artificial Intelligence, 2024, pp. 6979–6987

    2024

  16. [25]

    Profd: Prompt-guided feature disentangling for occluded person re- identification,

    C. Cui, S. Huang, W. Song, P. Ding, M. Zhang, and D. Wang, “Profd: Prompt-guided feature disentangling for occluded person re- identification,” inProceedings of the ACM International Conference on Multimedia, 2024, pp. 1583–1592

    2024

  17. [26]

    Region generation and assessment network for occluded person re- identification,

    S. He, W. Chen, K. Wang, H. Luo, F. Wang, W. Jiang, and H. Ding, “Region generation and assessment network for occluded person re- identification,”IEEE Transactions on Information Forensics and Secu- rity, vol. 19, pp. 120–132, 2024

    2024

  18. [27]

    Beyond appearance: A semantic controllable self-supervised learning framework for human-centric visual tasks,

    W. Chen, X. Xu, J. Jia, H. Luo, Y . Wang, F. Wang, R. Jin, and X. Sun, “Beyond appearance: A semantic controllable self-supervised learning framework for human-centric visual tasks,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp....

    2023

  19. [28]

    Pose-guided feature dis- entangling for occluded person re-identification based on transformer,

    T. Wang, H. Liu, P. Song, T. Guo, and W. Shi, “Pose-guided feature dis- entangling for occluded person re-identification based on transformer,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 36, no. 3, 2022, pp. 2540–2549

    2022

  20. [29]

    Keypoint-guided modality-invariant discriminative learning for visible-infrared person re- identification,

    T. Liang, Y . Jin, W. Liu, S. Feng, T. Wang, and Y . Li, “Keypoint-guided modality-invariant discriminative learning for visible-infrared person re- identification,” inProceedings of the ACM International Conference on Multimedia, 2022, pp. 3965–3973

    2022

  21. [30]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2016, pp. 770–778

    2016

  22. [31]

    An image is worth 16x16 words: Transformers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby, “An image is worth 16x16 words: Transformers for image recognition at scale,” inProceedings of the International Conf...

    2021

  23. [32]

    A convnet for the 2020s,

    Z. Liu, H. Mao, C. Wu, C. Feichtenhofer, T. Darrell, and S. Xie, “A convnet for the 2020s,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 11 966–11 976

    2022

  24. [33]

    Learning transferable visual models from natural language supervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agar- wal, G. Sastry, A. Askell, P. Mishkin, J. Clark, G. Krueger, and I. Sutskever, “Learning transferable visual models from natural language supervision,” inProceedings of the International conference on machine learn...

    2021

  25. [34]

    Learning to prompt for vision- language models,

    K. Zhou, J. Yang, C. C. Loy, and Z. Liu, “Learning to prompt for vision- language models,”International Journal of Computer Vision, vol. 130, no. 9, pp. 2337–2348, 2022

    2022

  26. [35]

    Conditional prompt learning for vision-language models,

    K. Zhou, J. Yang, C. C. Loy, and Z. Liu, “Conditional prompt learning for vision-language models,” inProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, 2022, pp. 16 795– 16 804

    2022

  27. [36]

    Unveiling the power of CLIP in unsupervised visible-infrared person re-identification,

    Z. Chen, Z. Zhang, X. Tan, Y . Qu, and Y . Xie, “Unveiling the power of CLIP in unsupervised visible-infrared person re-identification,” in Proceedings of the ACM International Conference on Multimedia, 2023, pp. 3667–3675

    2023

  28. [37]

    Visda: The visual domain adaptation challenge,

    X. Peng, B. Usman, N. Kaushik, J. Hoffman, D. Wang, and K. Saenko, “Visda: The visual domain adaptation challenge,”arXiv preprint arXiv:1710.06924, 2017

    2017 arXiv

  29. [38]

    Align your prompts: Test-time prompting with distribution alignment for zero-shot generalization,

    J. A. Samadh, H. Gani, N. Hussein, M. U. Khattak, M. Naseer, F. S. Khan, and S. H. Khan, “Align your prompts: Test-time prompting with distribution alignment for zero-shot generalization,” inProceedings of the Conference on Neural Information Processing Systems, 2023

    2023

  30. [39]

    Deep domain adaptation based multi-spectral salient object detection,

    S. Song, Z. Miao, H. Yu, J. Fang, K. Zheng, C. Ma, and S. Wang, “Deep domain adaptation based multi-spectral salient object detection,”IEEE Transactions on Multimedia, vol. 24, pp. 128–140, 2022

    2022

  31. [40]

    Incomplete multi- view clustering via prototype-based imputation,

    H. Li, Y . Li, M. Yang, P. Hu, D. Peng, and X. Peng, “Incomplete multi- view clustering via prototype-based imputation,” inProceedings of the International Joint Conference on Artificial Intelligence, 2023, pp. 3911– 3919

    2023

  32. [41]

    What makes training multi- modal classification networks hard?

    W. Wang, D. Tran, and M. Feiszli, “What makes training multi- modal classification networks hard?” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020, pp. 12 692–12 702. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 14

    2020

  33. [42]

    Balanced multimodal learning via on-the-fly gradient modulation,

    X. Peng, Y . Wei, A. Deng, D. Wang, and D. Hu, “Balanced multimodal learning via on-the-fly gradient modulation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 8228–8237

    2022

  34. [43]

    Adaptive transfer network for cross-domain person re-identification,

    J. Liu, Z. Zha, D. Chen, R. Hong, and M. Wang, “Adaptive transfer network for cross-domain person re-identification,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019, pp. 7202–7211

    2019

  35. [44]

    Camera topology graph guided vehicle re-identification,

    H. Li, A. Zheng, L. Sun, and Y . Luo, “Camera topology graph guided vehicle re-identification,”IEEE Transactions on Multimedia, vol. 26, pp. 1565–1577, 2024

    2024

  36. [45]

    Rgb-infrared cross- modality person re-identification,

    A. Wu, W. Zheng, H. Yu, S. Gong, and J. Lai, “Rgb-infrared cross- modality person re-identification,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2017, pp. 5390–5399

    2017

  37. [46]

    Progressive attribute embedding for accurate cross-modality person re-id,

    A. Zheng, P. Pan, H. Li, C. Li, B. Luo, C. Tan, and R. Jia, “Progressive attribute embedding for accurate cross-modality person re-id,” inPro- ceedings of the ACM International Conference on Multimedia, 2022, pp. 4309–4317

    2022

  38. [47]

    Cross-modality transformer with modality mining for visible-infrared person re-identification,

    T. Liang, Y . Jin, W. Liu, and Y . Li, “Cross-modality transformer with modality mining for visible-infrared person re-identification,”IEEE Transactions on Multimedia, vol. 25, pp. 8432–8444, 2023

    2023

  39. [48]

    Modality-adaptive mixup and invariant decomposition for rgb-infrared person re-identification,

    Z. Huang, J. Liu, L. Li, K. Zheng, and Z. Zha, “Modality-adaptive mixup and invariant decomposition for rgb-infrared person re-identification,” inProceedings of the International Joint Conference on Artificial Intelligence, 2022, pp. 1034–1042

    2022

  40. [49]

    Deep adversarial graph attention convolution network for text-based person search,

    J. Liu, Z. Zha, R. Hong, M. Wang, and Y . Zhang, “Deep adversarial graph attention convolution network for text-based person search,” in Proceedings of the ACM International Conference on Multimedia, 2019, pp. 665–673

    2019

  41. [50]

    Modality-aware style adaptation for rgb-infrared person re-identification,

    Z. Miao, H. Liu, W. Shi, W. Xu, and H. Ye, “Modality-aware style adaptation for rgb-infrared person re-identification,” inProceedings of the International Joint Conference on Artificial Intelligence, 2021, pp. 916–922

    2021

  42. [51]

    Visible- infrared person re-identification via cross-modality interaction trans- former,

    Y . Feng, J. Yu, F. Chen, Y . Ji, F. Wu, S. Liu, and X. Jing, “Visible- infrared person re-identification via cross-modality interaction trans- former,”IEEE Transactions on Multimedia, vol. 25, pp. 7647–7659, 2023

    2023

  43. [52]

    Diverse embedding expansion network and low-light cross-modality benchmark for visible-infrared person re- identification,

    Y . Zhang and H. Wang, “Diverse embedding expansion network and low-light cross-modality benchmark for visible-infrared person re- identification,” inProceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, 2023, pp. 2153–2162

    2023

  44. [53]

    A multi-constraint similarity learning with adaptive weighting for visible-thermal person re-identification,

    Y . Ling, Z. Luo, Y . Lin, and S. Li, “A multi-constraint similarity learning with adaptive weighting for visible-thermal person re-identification,” inProceedings of the International Joint Conference on Artificial Intelligence, 2021, pp. 845–851

    2021

  45. [54]

    Interact, embed, and enlarge: Boosting modality-specific representations for multi-modal person re-identification,

    Z. Wang, C. Li, A. Zheng, R. He, and J. Tang, “Interact, embed, and enlarge: Boosting modality-specific representations for multi-modal person re-identification,” inProceedings of the AAAI Conference on Artificial Intelligence, 2022, pp. 2633–2641

    2022

  46. [55]

    BLIP: bootstrapping language-image pre-training for unified vision-language understanding and generation,

    J. Li, D. Li, C. Xiong, and S. C. H. Hoi, “BLIP: bootstrapping language-image pre-training for unified vision-language understanding and generation,” inProceedings of the International conference on machine learning, vol. 162, 2022, pp. 12 888–12 900

    2022

  47. [56]

    Visual prompt tuning,

    M. Jia, L. Tang, B. Chen, C. Cardie, S. J. Belongie, B. Hariharan, and S. Lim, “Visual prompt tuning,” inProceedings of the European Conference on Computer Vision, vol. 13693, 2022, pp. 709–727

    2022

  48. [57]

    Adapt- former: Adapting vision transformers for scalable visual recognition,

    S. Chen, C. Ge, Z. Tong, J. Wang, Y . Song, J. Wang, and P. Luo, “Adapt- former: Adapting vision transformers for scalable visual recognition,” inProceedings of the Conference on Neural Information Processing Systems, vol. 35, 2022, pp. 16 664–16 678

    2022

  49. [58]

    Multi-scale deep learning architectures for person re-identification,

    X. Qian, Y . Fu, Y . Jiang, T. Xiang, and X. Xue, “Multi-scale deep learning architectures for person re-identification,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2017, pp. 5409–5418

    2017

  50. [59]

    Multi-level factorisation net for person re-identification,

    X. Chang, T. M. Hospedales, and T. Xiang, “Multi-level factorisation net for person re-identification,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018, pp. 2109–2118

    2018

  51. [60]

    Harmonious attention network for person re-identification,

    W. Li, X. Zhu, and S. Gong, “Harmonious attention network for person re-identification,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018, pp. 2285–2294

    2018

  52. [61]

    Counterfactual attention learning for fine-grained visual categorization and re-identification,

    Y . Rao, G. Chen, J. Lu, and J. Zhou, “Counterfactual attention learning for fine-grained visual categorization and re-identification,” inProceed- ings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 1005–1014

    2021

  53. [62]

    Unicat: Crafting a stronger fusion baseline for multimodal re-identification,

    J. Crawford, H. Yin, L. McDermott, and D. Cummings, “Unicat: Crafting a stronger fusion baseline for multimodal re-identification,” arXiv preprint arXiv:2310.18812, 2023

    2023 arXiv

  54. [63]

    Deep meta metric learning,

    G. Chen, T. Zhang, J. Lu, and J. Zhou, “Deep meta metric learning,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2019, pp. 9546–9555

    2019

  55. [64]

    Circle loss: A unified perspective of pair similarity optimization,

    Y . Sun, C. Cheng, Y . Zhang, C. Zhang, L. Zheng, Z. Wang, and Y . Wei, “Circle loss: A unified perspective of pair similarity optimization,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020, pp. 6397–6406

    2020

  56. [65]

    Heterogeneous rela- tional complement for vehicle re-identification,

    J. Zhao, Y . Zhao, J. Li, K. Yan, and Y . Tian, “Heterogeneous rela- tional complement for vehicle re-identification,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 205–214

    2021

  57. [66]

    Deep learning for person re-identification: A survey and outlook,

    M. Ye, J. Shen, G. Lin, T. Xiang, L. Shao, and S. C. H. Hoi, “Deep learning for person re-identification: A survey and outlook,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 6, pp. 2872–2893, 2022

    2022

  58. [67]

    Feature erasing and diffusion network for occluded person re-identification,

    Z. Wang, F. Zhu, S. Tang, R. Zhao, L. He, and J. Song, “Feature erasing and diffusion network for occluded person re-identification,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 4744–4753

    2022

  59. [68]

    Abd-net: Attentive but diverse person re-identification,

    T. Chen, S. Ding, J. Xie, Y . Yuan, W. Chen, Y . Yang, Z. Ren, and Z. Wang, “Abd-net: Attentive but diverse person re-identification,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019, pp. 8350–8360

    2019

  60. [69]

    Closing the domain gap for cross-modal visible-infrared vehicle re-identification,

    E. Kamenou, J. M. del Rinc ´on, P. Miller, and P. Devlin-Hill, “Closing the domain gap for cross-modal visible-infrared vehicle re-identification,” inProceedings of the International Conference on Pattern Recognition. IEEE, 2022, pp. 2728–2734

    2022

  61. [70]

    Generative and attentive fusion for multi-spectral vehicle re-identification,

    J. Guo, X. Zhang, Z. Liu, and Y . Wang, “Generative and attentive fusion for multi-spectral vehicle re-identification,” inProceedings of the Inter- national Conference on Intelligent Computing and Signal Processing, 2022, pp. 1565–1572

    2022

  62. [71]

    Graft: Gradual fusion transformer for multimodal re-identification,

    H. Yin, J. Li, E. Schiller, L. McDermott, and D. Cummings, “Graft: Gradual fusion transformer for multimodal re-identification,”arXiv preprint arXiv:2310.16856, 2023

    2023 arXiv

  63. [72]

    Graph-based progressive fusion network for multi-modality vehicle re-identification,

    Q. He, Z. Lu, Z. Wang, and H. Hu, “Graph-based progressive fusion network for multi-modality vehicle re-identification,”IEEE Transactions on Intelligent Transportation Systems, vol. 24, no. 11, pp. 12 431–12 447, 2023

    2023

  64. [73]

    Progressively hybrid transformer for multi-modal vehicle re-identification,

    W. Pan, L. Huang, J. Liang, L. Hong, and J. Zhu, “Progressively hybrid transformer for multi-modal vehicle re-identification,”Sensors, vol. 23, no. 9, p. 4206, 2023

    2023

  65. [74]

    Random erasing data augmentation,

    Z. Zhong, L. Zheng, G. Kang, S. Li, and Y . Yang, “Random erasing data augmentation,” inProceedings of the AAAI Conference on Artificial Intelligence, 2020, pp. 13 001–13 008

    2020

  66. [75]

    Grad-cam: Visual explanations from deep networks via gradient-based localization,

    R. R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, and D. Batra, “Grad-cam: Visual explanations from deep networks via gradient-based localization,” inProceedings of the IEEE/CVF Interna- tional Conference on Computer Vision, 2017, pp. 618–626

    2017

  67. [76]

    Visualizing data using t-sne,

    L. Van der Maaten and G. Hinton, “Visualizing data using t-sne,”Journal of machine learning research, vol. 9, no. 11, 2008

    2008

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.