Pith. sign in

REVIEW 4 major objections 5 minor 51 references

CP-DETR: Concept Prompt Guide DETR Toward Stronger Universal Object Detection

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A single pre-trained detector can set zero-shot records and match fully fine-tuned specialists by switching among text, visual, and optimized concept prompts.

desk verdict Solid engineering with a genuinely new combination—visual prompts inside an early-fusion DETR—but the headline SOTA claim over Grounding DINO is confounded by training data and unverifiable without code. read the letter →

arxiv 2412.09799 v1 pith:476RYVLU submitted 2024-12-13 cs.CV cs.AI

Qibo Chen , Weizhong Jin , Jianyue Ge , Mengdi Liu , Yuchao Yan , Jian Jiang , Li Yu , Xuanjiang Guo
show 2 more authors
Shuchang Li Jianzhong Chen
This is my paper · ORCID
classification cs.CVcs.AI
keywords universalobjectdetectionopen-vocabularyDETRconceptpromptsvisualprompttuningcross-modalfusionzero-shot
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

CP-DETR argues that the bottleneck in universal object detection is not dataset scale but how prompt information is fused with visual features. It builds a DETR-style detector whose encoder fuses concept prompts—vectors that represent object categories—with image features scale by scale, then gates across scales, so classification and localization are both conditioned on the prompt. With one pre-trained weight it reports 47.6 zero-shot AP on LVIS with Swin-T, 32.2 zero-shot AP on ODinW35 with Swin-L, 68.4 AP on COCO val through interactive visual prompts, and 73.1 AP on ODinW13 with optimized prompts. The point of the design is that one weight can switch among text, visual, and optimized prompts to remove alignment bias in downstream tasks.

What carries the argument

Concept prompts unify the three input modalities: text, box coordinates, and tuned embeddings all become high-dimensional vectors in one space. The load-bearing mechanism is the prompt visual hybrid encoder, made of progressive single-scale fusion (PSF) and multi-scale fusion gating (MFG): PSF starts from the deepest feature map so low-level features with weak semantics are not fused prematurely, while MFG flattens all scales and uses cross-modal attention plus deformable self-attention to gate information into the final concept prompts. A visual prompt encoder turns box coordinates into queries that aggregate concept vectors from multi-scale features, and an optimized prompt is a learnable embedding layer that represents each category with several vectors (ten by default) so that one category can absorb variation across datasets.

What would settle it

Train CP-DETR-T and Grounding DINO-T on the identical Objects365+GoldG data with no V3Det and the same schedule, and compare LVIS zero-shot AP; if the gap shrinks toward the 0.6 AP level of DetCLIPv3-T, the headline +20.2 AP is explained by data, not design.

Watch

Extended reading notes

Core claim

The paper's central claim is that effective cross-modal interaction is the missing ingredient in universal object detection: prior early-fusion models update all visual scales with prompts at once and ignore the semantic gaps between feature levels. CP-DETR's prompt visual hybrid encoder instead fuses the deepest feature map with prompts first, propagates that interaction top-down and bottom-up through single fusion layers, and then applies multi-scale fusion gating with deformable self-attention to gather critical information. Two auxiliary losses, a prompt multi-label loss and an anchor-based auxiliary detection head, give the hybrid encoder dense supervision during pre-training. The result, as reported, is a detector that with one fixed weight outperforms similarly sized universal detectors on zero-shot benchmarks and, when prompts are optimized, matches or beats fully fine-tuned specialist models.

Load-bearing premise

The load-bearing premise is that the reported benchmark gaps come from the proposed architecture rather than from the extra training data that only CP-DETR-T was given; if that data difference explains most of the gain, the architecture's superiority is not established.

Editorial extensions

If this is right

  • Universal detection gains can come from how prompts are fused rather than only from larger pre-training corpora: CP-DETR-T uses public O365, V3Det, and GoldG data and still beats DetCLIPv3-T, which adds 50M private captions, on LVIS zero-shot AP.
  • A single frozen weight can be specialized to a downstream dataset by optimizing only prompt embeddings, reaching 73.1 AP on ODinW13 and matching or exceeding full-model fine-tuning baselines.
  • Interactive detection with box visual prompts outperforms text prompts by 18.4 AP on ODinW35 for CP-DETR-L, giving users a practical way to remove alignment bias by pointing at examples.
  • The architecture keeps inference cost nearly flat as the number of categories grows (12.2 to 11.2 FPS at 1 to 80 classes for CP-DETR-T), whereas the main baseline's cost rises sharply.
  • Because the detector and the concept prompt generator are separate parts, a new prompt modality can be added without retraining the detector.

Reading between the lines

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

  • Beyond the paper: the +20.2 AP over Grounding DINO-T on LVIS is confounded by training data, since CP-DETR-T uses V3Det while the baseline does not; DetCLIPv3-T, which uses V3Det, trails by only 0.6 AP, so the architecture-specific gain is not yet isolated.
  • Beyond the paper: if prompt-only tuning closes most of the gap to full fine-tuning, then text-image alignment bias, not detector capacity, is the main barrier to downstream specialization, and other deep-fusion open-vocabulary detectors should show similar prompt-tuning gains.
  • Beyond the paper: the super-class representation may also absorb annotation noise and label ambiguity within a single dataset, which could be tested by increasing the number of prompt vectors per category on noisy labels.
  • Beyond the paper: the visual prompt encoder suggests that box-conditioned queries could transfer to open-vocabulary segmentation or referring-expression grounding with the same frozen detector.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes CP-DETR, a DETR-based universal object detector that conditions detection on concept prompts in a unified vector space. The detector uses a prompt-visual hybrid encoder built from progressive single-scale fusion (PSF) and multi-scale fusion gating (MFG), with auxiliary supervision from a prompt multi-label loss and an anchor-based detection head. Three concept-prompt generators are presented: text prompts, visual prompts, and optimized prompts via super-class prompt tuning. The authors report strong empirical results with a single pre-trained weight: 47.6 zero-shot AP on LVIS with Swin-T, 32.2 zero-shot AP on ODinW35 with Swin-L, 68.4 AP on COCO val for interactive visual prompting, and 73.1 AP on ODinW13 with optimized prompts. The paper also includes ablations, efficiency comparisons, and a limitation section.

Significance. If the reported results are controlled and reproducible, the paper makes a useful contribution to universal object detection: it demonstrates a computationally efficient cross-modal fusion design, introduces a unified concept-prompt formulation covering text, visual, and optimized prompts, and achieves competitive numbers without private pseudo-labeled data. The ablation study and the appendix's efficiency analysis are informative, and the paper is transparent about several limitations. However, the central state-of-the-art claim is weakened by training-data confounds in the headline comparisons, an ablation baseline that does not correspond to the main competitor's fusion module, and an unspecified ODinW35 evaluation protocol. The absence of release code or model weights further prevents independent verification. The significance is therefore conditional on addressing these control issues.

major comments (4)
  1. [Table 1, 'Comparison with Universal Detectors'] The headline LVIS comparison is confounded by training data: CP-DETR-T is pre-trained on O365+V3Det+GoldG, whereas Grounding DINO-T uses O365+GoldG+Cap4M and never sees V3Det's extra categories. The +20.2 AP difference on LVIS minival therefore mixes architectural gains with a large vocabulary/data difference. This is made more acute by the fact that DetCLIPv3-T, which also trains on V3Det plus additional private data, is only 0.6 AP behind CP-DETR-T. A data-controlled comparison is needed to support the claim of architectural superiority.
  2. [Table 4, 'Ablation'] The ablation row labeled 'replaced by DINO encoder' is not a control for Grounding DINO's cross-modal feature enhancement: it replaces the hybrid encoder with the closed-set DINO encoder, so the comparison measures early fusion against a non-cross-modal encoder rather than against the specific fusion design of the main competitor. The gains reported for the hybrid encoder (LVIS 42.2 to 44.3, ODinW13 58.5 to 64.0) are internally consistent but do not establish superiority over existing early-fusion encoders such as GLIP-style or Grounding DINO-style fusion under identical data.
  3. ['Evaluation Benchmark' and 'Comparison with Universal Detectors'] The paper states that ODinW35 evaluation 'followed the APE evaluation setup' without specifying what that setup changes relative to the standard ELEVATER protocol. If class-name corrections or resolution settings differ, the reported 32.2 AP is not directly comparable with the other numbers in Table 1. The exact modified protocol, including which categories were renamed and how images were resized, must be stated so the zero-shot ODinW35 claim can be verified.
  4. [Appendix, Figure 3, and Table 2] The super-class representation length M is selected by evaluating on ODinW13, which is the same benchmark used to report the headline 73.1 AP for optimized prompts. Tuning this hyperparameter on the evaluation benchmark can inflate the reported result; the paper should report the selection procedure, and ideally validate M on a held-out split or show sensitivity across benchmarks.
minor comments (5)
  1. ['Text Prompt Direct Evaluation'] The terms 'zero-shot' and 'full-shot' are used for Table 1, but the table does not explicitly mark which benchmarks were included in pre-training for each model; the gray/black color convention is helpful but should be stated in the caption and also applied to the LVIS columns for CP-DETR-L.
  2. [Equation (8)] The visual prompt training objective uses K and P_i^v / P_i^t, but the meaning of P_i is not fully defined; clarify that P_i is the i-th positive category's concept prompt and specify the alignment between P_v and P_t.
  3. [Equation (4)] The formula for P_end uses the dot product of P^{l+1} and P^l followed by Linear and ReLU, but the intended gating operation and normalization order are not entirely clear; adding a short explanation or a diagram of the tensor shapes would help.
  4. [Table 4] The row labels 'add row3 and super-class' are confusing because row 4 already includes row 3 plus the auxiliary head; rephrasing the rows as incremental additions would make the ablation easier to read.
  5. ['About Code'] The statement that code release requires internal permission is understandable, but given the empirical nature of the claims, the paper should at least provide a detailed evaluation configuration and, if possible, release feature visualizations or checkpoints through an institutional review process.

Circularity Check

1 steps flagged · score 2.0 of 10

Only minor circularity: the optimized-prompt ODinW13 record is tuned on that benchmark; the zero-shot claims remain independent external measurements.

  1. fitted input called prediction [Appendix, 'Additional Experiment' (Figure 3) and Experiments, Table 2]
    "As shown in figure 3, the performance on the downstream task gradually improves as the representation length increases, approaching saturation at 10, so we use 10 as the default length for optimized prompts."

    The same benchmark that anchors the headline optimized-prompt result (ODinW13, 73.1 AP in Table 2) is also the benchmark used to select the super-class representation length. Figure 3 ablates the representation length on ODinW13 full-shot AP and chooses M=10 based on that metric; the selected configuration is then reported as the optimized-prompt record. The reported 73.1 AP is therefore not an independent evaluation of a fixed method: a hyperparameter of the proposed 'super-class representation' was chosen by optimizing the very benchmark on which the method is then declared state-of-the-art. This is a mild in-sample selection effect, not a derivation from first principles, and it does not affect the zero-shot text-prompt results.

full rationale

CP-DETR's central zero-shot claims (47.6 AP on LVIS, 32.2 AP on ODinW35) are external benchmark measurements with a fixed pre-trained weight; they are not derived from any fitted parameter of the paper, and no equation reduces a claimed prediction to its own input. The only circular step I can exhibit is the super-class representation length for optimized prompts: the appendix ablates this length on ODinW13 and selects 10 on the basis of full-shot AP on that benchmark, and the same ODinW13 benchmark then anchors the headline 73.1 AP record in Table 2. That number is therefore in-sample with respect to hyperparameter selection and is not an independent confirmation of the method; however, it affects only the fully-shot optimized-prompt claim and does not infect the zero-shot comparisons. The remaining concerns—uncontrolled V3Det training data versus Grounding DINO, the 'DINO encoder' ablation row, and the unspecified APE-modified ODinW35 setup—are benchmark-control and correctness issues, not circularity. The paper's self-citation to Chen et al. 2024 is used only as a related baseline/prompt-learning reference and is not load-bearing. No uniqueness theorem, imported ansatz, or renaming of a known result is present.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claim is empirical, so this ledger lists the design hyperparameters and background assumptions that the reported performance depends on. No new physical entities are introduced.

free parameters (5)
  • Super-class representation length M = 10 (per category, by ablation on ODinW13)
    Number of learned prompt vectors per class for optimized prompts; chosen by sweeping values 1, 2, 3, 5, 10, 15, 20 on ODinW13 (Figure 3).
  • Number of negative text samples per batch = 80
    Randomly sampled negative categories used to reduce detection hallucination; chosen without ablation.
  • Memory bank length for text dictionary = 1000
    Online text dictionary size for grounding and REC data; set by hand.
  • Auxiliary loss weights = prompt multi-label 6, aux class 6, centerness 6, IoU 12
    Loss balancing weights for auxiliary supervision, set following prior work.
  • Training data composition and sampling ratios = see Table 6
    Sampling ratios for O365, V3Det, GoldG, OI, VG, RefC, COCO, LVIS, PhraseCut affect performance and are chosen by the authors.
assumptions (5)
  • domain assumption CLIP/EVA-02 text encoder provides a semantic space that supports open-set concept generalization.
    Used as the text prompt encoder; if the text space is poorly aligned, zero-shot claims fail.
  • domain assumption DETR with language-guided query selection and cross-modality decoder provides a sound detection backbone.
    The detector builds on Grounding DINO and DINO and does not re-derive these components.
  • domain assumption Public datasets with region-text annotations are sufficient for universal detection pre-training.
    The model is trained jointly on O365, V3Det, GoldG, etc.; no private data is used, so claims rest on their quality and coverage.
  • ad hoc to paper Auxiliary supervision (multi-label loss and anchor head) improves the hybrid encoder's fusion learning.
    The paper argues this, but the ablation support is modest (rows 3-4 in Table 4, +0.5 AP on LVIS).
  • ad hoc to paper The APE-adjusted ODinW35 evaluation setup is a fair measure of zero-shot performance.
    The paper drops or fixes datasets with 'quality issues' following APE; this affects the headline 32.2 AP number.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CP-DETR: Concept Prompt Guide DETR Toward Stronger Universal Object Detection." pith.science (2026). https://pith.science/paper/476RYVLU

@misc{pith2026241209799,
  author       = {Pith},
  title        = {Pith review of: CP-DETR: Concept Prompt Guide DETR Toward Stronger Universal Object Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/476RYVLU}},
  note         = {Machine review of arXiv:2412.09799}
}
read the original abstract

Recent research on universal object detection aims to introduce language in a SoTA closed-set detector and then generalize the open-set concepts by constructing large-scale (text-region) datasets for training. However, these methods face two main challenges: (i) how to efficiently use the prior information in the prompts to genericise objects and (ii) how to reduce alignment bias in the downstream tasks, both leading to sub-optimal performance in some scenarios beyond pre-training. To address these challenges, we propose a strong universal detection foundation model called CP-DETR, which is competitive in almost all scenarios, with only one pre-training weight. Specifically, we design an efficient prompt visual hybrid encoder that enhances the information interaction between prompt and visual through scale-by-scale and multi-scale fusion modules. Then, the hybrid encoder is facilitated to fully utilize the prompted information by prompt multi-label loss and auxiliary detection head. In addition to text prompts, we have designed two practical concept prompt generation methods, visual prompt and optimized prompt, to extract abstract concepts through concrete visual examples and stably reduce alignment bias in downstream tasks. With these effective designs, CP-DETR demonstrates superior universal detection performance in a broad spectrum of scenarios. For example, our Swin-T backbone model achieves 47.6 zero-shot AP on LVIS, and the Swin-L backbone model achieves 32.2 zero-shot AP on ODinW35. Furthermore, our visual prompt generation method achieves 68.4 AP on COCO val by interactive detection, and the optimized prompt achieves 73.1 fully-shot AP on ODinW13.

Figures

Figures reproduced from arXiv: 2412.09799 by the authors.

Figure 1
Figure 1. Overall architecture of CP-DETR. First, the concept prompt generator (shown in green dashed box) encodes textual [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. The overall architecture of the visual prompt en [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Ablation results for the super-class representation [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Visualizations of CP-DETR-L zero-shot outputs. [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: Visualizations of CP-DETR-L visual prompt outputs. Row 1 use of a class of boxes as inputs. Row 2 use of two [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 34 canonical work pages

  1. [1]

    Chen, K.; Wang, J.; Pang, J.; Cao, Y.; Xiong, Y.; Li, X.; Sun, S.; Feng, W.; Liu, Z.; Xu, J.; et al. 2019. MMDetection: Open mmlab detection toolbox and benchmark. arXiv:1906.07155

    arXiv 2019

  2. [2]

    Chen, Q.; Jin, W.; Li, S.; Liu, M.; Yu, L.; Jiang, J.; and Wang, X. 2024. Exploration of visual prompt in Grounded pre-trained open-set detection. In ICASSP, 6115--6119. IEEE

    work page 2024

  3. [3]

    Cheng, T.; Song, L.; Ge, Y.; Liu, W.; Wang, X.; and Shan, Y. 2024. Yolo-world: Real-time open-vocabulary object detection. In CVPR, 16901--16911

    work page 2024

  4. [4]

    Devlin, J.; Chang, M.-W.; Lee, K.; and Toutanova, K. 2018. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv:1810.04805

    arXiv 2018

  5. [5]

    Ding, X.; Zhang, X.; Ma, N.; Han, J.; Ding, G.; and Sun, J. 2021. Repvgg: Making vgg-style convnets great again. In CVPR, 13733--13742

    work page 2021

  6. [6]

    Fang, Y.; Sun, Q.; Wang, X.; Huang, T.; Wang, X.; and Cao, Y. 2024. Eva-02: A visual representation for neon genesis. Image and Vision Computing, 105171

    work page 2024

  7. [7]

    Feng, C.; Zhong, Y.; Jie, Z.; Chu, X.; Ren, H.; Wei, X.; Xie, W.; and Ma, L. 2022. Promptdet: Towards open-vocabulary detection using uncurated images. In ECCV, 701--717. Springer

    work page 2022

  8. [8]

    Gu, X.; Lin, T.-Y.; Kuo, W.; and Cui, Y. 2022. Open-vocabulary Object Detection via Vision and Language Knowledge Distillation. In ICLR

    2022

Show all 51 references
  1. [9]

    Gupta, A.; Dollar, P.; and Girshick, R. 2019. Lvis: A dataset for large vocabulary instance segmentation. In CVPR, 5356--5364

    2019

  2. [10]

    Jiang, Q.; Li, F.; Zeng, Z.; Ren, T.; Liu, S.; and Zhang, L. 2024. T-Rex2: Towards Generic Object Detection via Text-Visual Prompt Synergy. arXiv:2403.14610

    2024 arXiv

  3. [11]

    Kamath, A.; Singh, M.; LeCun, Y.; Synnaeve, G.; Misra, I.; and Carion, N. 2021. Mdetr-modulated detection for end-to-end multi-modal understanding. In ICCV, 1780--1790

    2021

  4. [12]

    A.; et al

    Krishna, R.; Zhu, Y.; Groth, O.; Johnson, J.; Hata, K.; Kravitz, J.; Chen, S.; Kalantidis, Y.; Li, L.-J.; Shamma, D. A.; et al. 2017. Visual genome: Connecting language and vision using crowdsourced dense image annotations. International journal of computer vision, 123: 32--73

    2017

  5. [13]

    Kuznetsova, A.; Rom, H.; Alldrin, N.; Uijlings, J.; Krasin, I.; Pont-Tuset, J.; Kamali, S.; Popov, S.; Malloci, M.; Kolesnikov, A.; et al. 2020. The open images dataset v4: Unified image classification, object detection, and visual relationship detection at scale. Internationa...

    2020

  6. [14]

    J.; and Gao, J

    Li, C.; Liu, H.; Li, L.; Zhang, P.; Aneja, J.; Yang, J.; Jin, P.; Hu, H.; Liu, Z.; Lee, Y. J.; and Gao, J. 2022 a . ELEVATER: A Benchmark and Toolkit for Evaluating Language-Augmented Visual Models. In NeurIPS, 9287--9301

    2022

  7. [15]

    H.; Zhang, P.; Zhang, H.; Yang, J.; Li, C.; Zhong, Y.; Wang, L.; Yuan, L.; Zhang, L.; Hwang, J.-N.; et al

    Li, L. H.; Zhang, P.; Zhang, H.; Yang, J.; Li, C.; Zhong, Y.; Wang, L.; Yuan, L.; Zhang, L.; Hwang, J.-N.; et al. 2022 b . Grounded language-image pre-training. In CVPR, 10965--10975

    2022

  8. [16]

    Li, X.; Wang, W.; Wu, L.; Chen, S.; Hu, X.; Li, J.; Tang, J.; and Yang, J. 2020. Generalized Focal Loss: Learning Qualified and Distributed Bounding Boxes for Dense Object Detection. In NeurIPS, 21002--21012

    2020

  9. [17]

    Li, Z.; Yao, L.; Zhang, X.; Wang, X.; Kanhere, S.; and Zhang, H. 2019. Zero-shot object detection with textual descriptions. In AAAI, 8690--8697

    2019

  10. [18]

    Lin, T.-Y.; Maire, M.; Belongie, S.; Hays, J.; Perona, P.; Ramanan, D.; Doll \'a r, P.; and Zitnick, C. L. 2014. Microsoft coco: Common objects in context. In ECCV, 740--755. Springer

    2014

  11. [19]

    Liu, S.; Qi, L.; Qin, H.; Shi, J.; and Jia, J. 2018. Path aggregation network for instance segmentation. In CVPR, 8759--8768

    2018

  12. [20]

    Liu, S.; Zeng, Z.; Ren, T.; Li, F.; Zhang, H.; Yang, J.; Li, C.; Yang, J.; Su, H.; Zhu, J.; et al. 2023. Grounding dino: Marrying dino with grounded pre-training for open-set object detection. arXiv:2303.05499

    2023 arXiv

  13. [21]

    Liu, Z.; Lin, Y.; Cao, Y.; Hu, H.; Wei, Y.; Zhang, Z.; Lin, S.; and Guo, B. 2021. Swin transformer: Hierarchical vision transformer using shifted windows. In ICCV, 10012--10022

    2021

  14. [22]

    L.; and Murphy, K

    Mao, J.; Huang, J.; Toshev, A.; Camburu, O.; Yuille, A. L.; and Murphy, K. 2016. Generation and comprehension of unambiguous object descriptions. In CVPR, 11--20

    2016

  15. [23]

    Minderer, M.; Gritsenko, A.; Houlsby, N.; et al. 2023. Scaling Open-Vocabulary Object Detection. In NeurIPS, 72983--73007

    2023

  16. [24]

    Minderer, M.; Gritsenko, A.; Stone, A.; Neumann, M.; Weissenborn, D.; Dosovitskiy, A.; Mahendran, A.; Arnab, A.; Dehghani, M.; Shen, Z.; et al. 2022. Simple open-vocabulary object detection. In ECCV, 728--755. Springer

    2022

  17. [25]

    W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al

    Radford, A.; Kim, J. W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al. 2021. Learning transferable visual models from natural language supervision. In ICML, 8748--8763. PMLR

    2021

  18. [26]

    Rezatofighi, H.; Tsoi, N.; Gwak, J.; Sadeghian, A.; Reid, I.; and Savarese, S. 2019. Generalized intersection over union: A metric and a loss for bounding box regression. In CVPR, 658--666

    2019

  19. [27]

    Shao, S.; Li, Z.; Zhang, T.; Peng, C.; Yu, G.; Zhang, X.; Li, J.; and Sun, J. 2019. Objects365: A large-scale, high-quality dataset for object detection. In ICCV, 8430--8439

    2019

  20. [28]

    Shen, Y.; Fu, C.; Chen, P.; Zhang, M.; Li, K.; Sun, X.; Wu, Y.; Lin, S.; and Ji, R. 2024. Aligning and prompting everything all at once for universal visual perception. In CVPR, 13193--13203

    2024

  21. [29]

    Wang, J.; Zhang, P.; Chu, T.; Cao, Y.; Zhou, Y.; Wu, T.; Wang, B.; He, C.; and Lin, D. 2023. V3det: Vast vocabulary visual detection dataset. In ICCV, 19844--19854

    2023

  22. [30]

    Wu, C.; Lin, Z.; Cohen, S.; Bui, T.; and Maji, S. 2020. Phrasecut: Language-based image segmentation in the wild. In CVPR, 10216--10225

    2020

  23. [31]

    Wu, J.; Jiang, Y.; Liu, Q.; Yuan, Z.; Bai, X.; and Bai, S. 2024. General object foundation model for images and videos at scale. In CVPR, 3783--3795

    2024

  24. [32]

    Xiao, B.; Wu, H.; Xu, W.; Dai, X.; Hu, H.; Lu, Y.; Zeng, M.; Liu, C.; and Yuan, L. 2024. Florence-2: Advancing a unified representation for a variety of vision tasks. In CVPR, 4818--4829

    2024

  25. [33]

    Xu, Y.; Zhang, M.; Fu, C.; Chen, P.; Yang, X.; Li, K.; and Xu, C. 2023. Multi-modal queried object detection in the wild. In NeurIPS, 4452--4469

    2023

  26. [34]

    Yan, B.; Jiang, Y.; Wu, J.; Wang, D.; Luo, P.; Yuan, Z.; and Lu, H. 2023. Universal instance perception as object discovery and retrieval. In CVPR, 15325--15336

    2023

  27. [35]

    Yao, L.; Han, J.; Liang, X.; Xu, D.; Zhang, W.; Li, Z.; and Xu, H. 2023. Detclipv2: Scalable open-vocabulary object detection pre-training via word-region alignment. In CVPR, 23497--23506

    2023

  28. [36]

    Yao, L.; Han, J.; Wen, Y.; Liang, X.; Xu, D.; Zhang, W.; Li, Z.; XU, C.; and Xu, H. 2022. DetCLIP: Dictionary-Enriched Visual-Concept Paralleled Pre-training for Open-world Detection. In NeurIPS, 9125--9138

    2022

  29. [37]

    Yao, L.; Pi, R.; Han, J.; Liang, X.; Xu, H.; Zhang, W.; Li, Z.; and Xu, D. 2024. DetCLIPv3: Towards Versatile Generative Open-vocabulary Object Detection. In CVPR, 27391--27401

    2024

  30. [38]

    C.; and Berg, T

    Yu, L.; Poirson, P.; Yang, S.; Berg, A. C.; and Berg, T. L. 2016. Modeling context in referring expressions. In ECCV, 69--85. Springer

    2016

  31. [39]

    Zang, Y.; Li, W.; Zhou, K.; Huang, C.; and Loy, C. C. 2022. Open-vocabulary detr with conditional matching. In ECCV, 106--122. Springer

    2022

  32. [40]

    Zhang, H.; Li, F.; Liu, S.; Zhang, L.; Su, H.; Zhu, J.; Ni, L.; and Shum, H.-Y. 2023. DINO : DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection. In ICLR

    2023

  33. [41]

    Zhang, H.; Zhang, P.; Hu, X.; Chen, Y.-C.; Li, L.; Dai, X.; Wang, L.; Yuan, L.; Hwang, J.-N.; and Gao, J. 2022. GLIPv2: Unifying Localization and Vision-Language Understanding. In NeurIPS, 36067--36080

    2022

  34. [42]

    Zhang, S.; Chi, C.; Yao, Y.; Lei, Z.; and Li, S. Z. 2020. Bridging the gap between anchor-based and anchor-free detection via adaptive training sample selection. In CVPR, 9759--9768

    2020

  35. [43]

    Zhao, T.; Liu, P.; He, X.; Zhang, L.; and Lee, K. 2024 a . Real-time Transformer-based Open-Vocabulary Detection with Efficient Fusion Head. arXiv:2403.06892

    2024 arXiv

  36. [44]

    Zhao, T.; Liu, P.; Lu, X.; and Lee, K. 2022. Omdet: Language-aware object detection with large-scale vision-language multi-dataset pre-training. arXiv:2209.05946

    2022 arXiv

  37. [45]

    Zhao, Y.; Lv, W.; Xu, S.; Wei, J.; Wang, G.; Dang, Q.; Liu, Y.; and Chen, J. 2024 b . Detrs beat yolos on real-time object detection. In CVPR, 16965--16974

    2024

  38. [46]

    H.; Zhou, L.; Dai, X.; Yuan, L.; Li, Y.; et al

    Zhong, Y.; Yang, J.; Zhang, P.; Li, C.; Codella, N.; Li, L. H.; Zhou, L.; Dai, X.; Yuan, L.; Li, Y.; et al. 2022. Regionclip: Region-based language-image pretraining. In CVPR, 16793--16803

    2022

  39. [47]

    C.; and Liu, Z

    Zhou, K.; Yang, J.; Loy, C. C.; and Liu, Z. 2022. Learning to prompt for vision-language models. International Journal of Computer Vision, 130(9): 2337--2348

    2022

  40. [48]

    Zhu, X.; Su, W.; Lu, L.; Li, B.; Wang, X.; and Dai, J. 2021. Deformable DETR: Deformable Transformers for End-to-End Object Detection. In ICLR

    2021

  41. [49]

    Zong, Z.; Song, G.; and Liu, Y. 2023. Detrs with collaborative hybrid assignments training. In ICCV, 6748--6758

    2023

  42. [50]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all...

  43. [51]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.