Pith. sign in

REVIEW 2 major objections 2 minor 57 references

AdaCount: Training-Free Similarity-Guided Spatial and Feature Adaptation for Zero-Shot Object Counting

T0 review · 2 major / 2 minor · reviewed 2026-07-03 · grok-4.3

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

Pith's one-line read AdaCount uses a text-prompt similarity map to warp images spatially and modulate features for improved zero-shot counting with SAM3.

desk verdict AdaCount adds similarity-map-guided warping and modulation to SAM3 for training-free ZOC, but the map's accuracy on dense small-object scenes is unverified and central to the claimed gains. read the letter →

arxiv 2607.02139 v1 pith:XM73K62M submitted 2026-07-02 cs.CV

Muhammad Ibraheem Siddiqui , Muhammad Haris Khan This is my paper · ORCID
classification cs.CV
keywords zero-shotobjectcountingtraining-freesimilarity-guidedadaptationspatialwarpingfeaturemodulationtextpromptSAM3
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 a prototype-driven similarity map derived from the text prompt can identify target regions and then direct two adaptations on SAM3: spatial warping to shift resolution toward those regions and feature modulation to boost relevant encoder outputs. This combination lets the model handle dense scenes with many small objects more effectively while keeping the full image context, all without any retraining or counting-specific data. A sympathetic reader would care because zero-shot counting from text alone becomes more reliable on existing foundation models across varied benchmarks where prior prompt-driven methods miss instances or fail to separate them.

What carries the argument

The prototype-driven similarity map from the text prompt that identifies target regions and guides both spatial warping and feature modulation.

What would settle it

Apply the similarity-guided adaptations to SAM3 on a dense small-object counting benchmark and measure whether missed instances or false positives increase compared with the unadapted baseline.

Watch

Extended reading notes

Core claim

AdaCount first estimates a prototype-driven similarity map that identifies target-relevant regions. This similarity map subsequently guides two complementary adaptations: (i) similarity-guided spatial warping, which reallocates image resolution toward target instances, and (ii) feature modulation, which amplifies target-relevant encoder representations. Together, these adaptations enable SAM3 to devote greater representational capacity to target-relevant regions while preserving global image context, without requiring any model retraining.

Load-bearing premise

The prototype-driven similarity map computed from the text prompt accurately identifies target-relevant regions even in densely populated scenes with numerous small objects.

Editorial extensions

If this is right

  • Better instance separation occurs in crowded scenes because resolution is reallocated toward target objects.
  • New state-of-the-art results among training-free zero-shot counting methods hold across six diverse benchmarks.
  • Global image context remains available while local focus on relevant regions increases.
  • No counting-specific training data or retraining is needed for the gains.

Reading between the lines

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

  • The same similarity-guided warping and modulation steps could be tested on other prompt-driven segmentation tasks that struggle with small or dense instances.
  • If the similarity map proves reliable, it might reduce the need for higher-resolution inputs in related dense-prediction problems.
  • Combining the map with refined text prompts could further lower error rates on categories where the initial prototype match is weak.
Share X Bluesky LinkedIn Reddit HN

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 / 2 minor

Summary. The paper introduces AdaCount, a training-free zero-shot object counting method that computes a prototype-driven similarity map from a text prompt and SAM3 encoder features, then uses this map to perform similarity-guided spatial warping (to reallocate resolution toward target regions) and feature modulation (to amplify target-relevant representations) before feeding the adapted input to SAM3 for instance segmentation and counting. It claims this addresses SAM3's failures on dense scenes with small objects and establishes new SOTA results among training-free ZOC methods on six benchmarks.

Significance. If the empirical gains are shown to be causally attributable to the proposed adaptations rather than post-hoc tuning or dataset-specific effects, the work would be significant for demonstrating how lightweight, training-free guidance from foundation-model features can mitigate resolution and attention limitations in promptable segmentation models without retraining or additional supervision.

major comments (2)
  1. [§3] §3 (similarity map computation and downstream use): the central claim that the prototype-driven similarity map reliably identifies target instances in densely populated scenes with small objects is load-bearing for both the spatial warping and feature modulation steps, yet the manuscript provides no quantitative validation of map quality (e.g., IoU or precision against point/mask annotations on the hardest subsets of the six benchmarks); without this, the reported counting improvements cannot be confidently attributed to the adaptations rather than other factors.
  2. [Experiments] Experiments section (SOTA claim): the assertion of new state-of-the-art among training-free ZOC approaches requires explicit comparison tables with all recent baselines (including SAM3 itself) plus ablations isolating the contribution of warping versus modulation; the absence of such breakdowns on the dense-small-object subsets undermines the causal link asserted in the abstract.
minor comments (2)
  1. [Abstract] Abstract: states SOTA performance on six benchmarks but supplies no numerical results, baseline names, or error metrics; while common, this makes the strength of the claim difficult to assess from the front matter alone.
  2. [§3] Notation in §3: the definition of the prototype vector and the exact similarity computation (cosine vs. other) should be written as an explicit equation to allow reproduction.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We respond point-by-point to the major comments and commit to revisions that directly address the concerns about causal attribution and experimental completeness.

read point-by-point responses

  1. Referee: [§3] §3 (similarity map computation and downstream use): the central claim that the prototype-driven similarity map reliably identifies target instances in densely populated scenes with small objects is load-bearing for both the spatial warping and feature modulation steps, yet the manuscript provides no quantitative validation of map quality (e.g., IoU or precision against point/mask annotations on the hardest subsets of the six benchmarks); without this, the reported counting improvements cannot be confidently attributed to the adaptations rather than other factors.

    Authors: We agree that direct quantitative validation of the similarity map would strengthen the causal link between the map and the observed counting gains. While end-to-end results and component ablations provide supporting evidence, we will add IoU, precision, and recall metrics for the similarity map against ground-truth annotations on the dense-small-object subsets of the benchmarks in the revised manuscript. revision: yes

  2. Referee: [Experiments] Experiments section (SOTA claim): the assertion of new state-of-the-art among training-free ZOC approaches requires explicit comparison tables with all recent baselines (including SAM3 itself) plus ablations isolating the contribution of warping versus modulation; the absence of such breakdowns on the dense-small-object subsets undermines the causal link asserted in the abstract.

    Authors: We acknowledge the value of more granular experimental breakdowns. The manuscript already reports comparisons against training-free ZOC baselines and SAM3, but we will expand the tables to make all comparisons fully explicit, add ablations that isolate spatial warping from feature modulation, and report these results specifically on the dense-small-object subsets to better substantiate the SOTA claim. revision: yes

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: similarity map computed from prompt/image then used for adaptations; empirical SOTA claim independent of derivation

full rationale

The paper describes a training-free pipeline in which a prototype-driven similarity map is first estimated from the text prompt and input image, after which the map is applied to guide spatial warping and feature modulation before feeding into SAM3. No equation or step is shown to reduce to its own inputs by construction, no fitted parameter is relabeled as a prediction, and no load-bearing premise rests on a self-citation chain. The central claim is an empirical performance improvement across benchmarks rather than a mathematical derivation, so the method remains self-contained against external evaluation.

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

Only the abstract is available; it does not enumerate free parameters, background axioms, or new entities. The method implicitly assumes access to a pre-trained SAM3 model and the ability to compute a reliable prototype similarity map from text and image features.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AdaCount: Training-Free Similarity-Guided Spatial and Feature Adaptation for Zero-Shot Object Counting." pith.science (2026). https://pith.science/paper/XM73K62M

@misc{pith2026260702139,
  author       = {Pith},
  title        = {Pith review of: AdaCount: Training-Free Similarity-Guided Spatial and Feature Adaptation for Zero-Shot Object Counting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XM73K62M}},
  note         = {Machine review of arXiv:2607.02139}
}
read the original abstract

Zero-shot object counting (ZOC) aims to count instances of arbitrary object categories specified only through textual prompts. Recent training-free approaches leverage foundation models such as SAM to reformulate counting as a prompt-driven segmentation task, eliminating the need for costly counting-specific training data with point-level annotations. More recently, SAM3 introduced promptable concept segmentation, enabling the zero-shot segmentation of all instances corresponding to a text-defined concept. However, SAM3 struggles in densely populated scenes containing numerous small objects, where limited image resolution and insufficient attention to target-relevant regions often lead to missed instances and poor instance separation, hindering accurate object counting. To address this limitation, we propose AdaCount, a training-free framework for ZOC based on similarity-guided spatial and feature adaptation. AdaCount first estimates a prototype-driven similarity map that identifies target-relevant regions. This similarity map subsequently guides two complementary adaptations: (i) similarity-guided spatial warping, which reallocates image resolution toward target instances, and (ii) feature modulation, which amplifies target-relevant encoder representations. Together, these adaptations enable SAM3 to devote greater representational capacity to target-relevant regions while preserving global image context, without requiring any model retraining. Extensive experiments across six diverse counting benchmarks establish AdaCount as a new SOTA among training-free ZOC approaches.

Figures

Figures reproduced from arXiv: 2607.02139 by the authors.

Figure 1
Figure 1. Comparison of SAM3, SAM3Count, and AdaCount [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of AdaCount: Given an input image and text query, AdaCount first performs an initial SAM3 pass to discover [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Attention-guided versus prototype-driven similarity [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Comparison of SAM3, SAM3Count, and AdaCount. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Attention map produces coarse region-level responses [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Qualitative comparison of SAM3 and AdaCount. From left to right: (1) input image and text prompt; (2) predictions produced by the vanilla SAM3 baseline; (3) the prototype-driven similarity map estimated from the initial SAM3 inference pass, highlighting target-relevant…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 57 canonical work pages

  1. [1]

    Open-world text-specifed object counting

    Niki Amini, Kiana Amini-Naieni, Tengda Han, and Andrew Zisserman. Open-world text-specifed object counting. In BMVC, 2023. 2, 6

    work page 2023

  2. [2]

    Countgd: Multi-modal open-world counting.Advances in Neural In- formation Processing Systems, 37:48810–48837, 2024

    Niki Amini, Tengda Han, and Andrew Zisserman. Countgd: Multi-modal open-world counting.Advances in Neural In- formation Processing Systems, 37:48810–48837, 2024. 2, 6

    work page 2024

  3. [3]

    Countgd++: Gen- eralized prompting for open-world counting

    Niki Amini-Naieni and Andrew Zisserman. Countgd++: Gen- eralized prompting for open-world counting. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 37725–37734, 2026. 6

    work page 2026

  4. [4]

    Open-world ob- ject counting in videos

    Niki Amini-Naieni and Andrew Zisserman. Open-world ob- ject counting in videos. InProceedings of the AAAI Con- ference on Artificial Intelligence, pages 2300–2308, 2026. 6

    work page 2026

  5. [5]

    Completely self-supervised crowd counting via distribution matching

    Deepak Babu Sam, Abhinav Agarwalla, Jimmy Joseph, Vish- wanath A Sindagi, R Venkatesh Babu, and Vishal M Patel. Completely self-supervised crowd counting via distribution matching. InEuropean Conference on Computer Vision, pages 186–204. Springer, 2022. 1

    work page 2022

  6. [6]

    Unveiling visual perception in language models: An attention head analysis approach

    Jing Bi, Junjia Guo, Yunlong Tang, Lianggong Bruce Wen, Zhang Liu, Bingjie Wang, and Chenliang Xu. Unveiling visual perception in language models: An attention head analysis approach. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 4135–4144, 2025. 3

    work page 2025

  7. [7]

    Perception encoder: The best visual embeddings are not at the output of the net- work.Advances in Neural Information Processing Systems, 38:60884–60937, 2026

    Daniel Bolya, Po-Yao Huang, Peize Sun, Jang Hyun Cho, Andrea Madotto, Chen Wei, Tengyu Ma, Jiale Zhi, Jathushan Rajasegaran, Hanoona Bangalath, et al. Perception encoder: The best visual embeddings are not at the output of the net- work.Advances in Neural Information Processing Systems, 38:60884–60937, 2026. 3

    work page 2026

  8. [8]

    A vehicle counts by class framework using distinguished regions tracking at mul- tiple intersections

    Nam Bui, Hongsuk Yi, and Jiho Cho. A vehicle counts by class framework using distinguished regions tracking at mul- tiple intersections. InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition Workshops, pages 578–579, 2020. 1

    work page 2020

Show all 57 references
  1. [9]

    Sam 3: Segment anything with concepts.The Fourteenth International Conference on Learning Representations., 2026

    Nicolas Carion, Laura Gustafson, Yuan-Ting Hu, Shoub- hik Debnath, Ronghang Hu, Didac Suris, Chaitanya Ryali, Kalyan Vasudev Alwala, Haitham Khedr, Andrew Huang, et al. Sam 3: Segment anything with concepts.The Fourteenth International Conference on Learning Representations., ...

    2026

  2. [10]

    Mind the prompt: A novel benchmark for prompt-based class-agnostic counting

    Luca Ciampi, Nicola Messina, Matteo Pierucci, Giuseppe Amato, Marco Avvenuti, and Fabrizio Falchi. Mind the prompt: A novel benchmark for prompt-based class-agnostic counting. In2025 IEEE/CVF Winter Conference on Applica- tions of Computer Vision (WACV), pages 7970–7979. IEEE,

  3. [11]

    Constructive distortion: Improving MLLMs with attention-guided image warping

    Dwip Dalal, Gautam Vashishtha, Utkarsh Mishra, Jeonghwan Kim, Madhav Kanda, Hyeonjeong Ha, Svetlana Lazebnik, Heng Ji, and Unnat Jain. Constructive distortion: Improving MLLMs with attention-guided image warping. InThe Four- teenth International Conference on Learning Representations,

  4. [12]

    Afreeca: Annotation-free counting for all

    Adriano D’Alessandro, Ali Mahdavi-Amiri, and Ghassan Hamarneh. Afreeca: Annotation-free counting for all. In European Conference on Computer Vision, pages 75–91. Springer, 2024. 1

    2024

  5. [13]

    Image retargeting using mesh parametrization

    Yanwen Guo, Feng Liu, Jian Shi, Zhi-Hua Zhou, and Michael Gleicher. Image retargeting using mesh parametrization. IEEE Transactions on Multimedia, 11(5):856–867, 2009. 3

    2009

  6. [14]

    Few-shot object counting with dynamic similarity-aware in latent space.IEEE Transactions on Geoscience and Remote Sensing, 62:1–14, 2024

    Jinghui He, Bo Liu, Fan Cao, Jian Xu, and Yanshan Xiao. Few-shot object counting with dynamic similarity-aware in latent space.IEEE Transactions on Geoscience and Remote Sensing, 62:1–14, 2024. 1

    2024

  7. [15]

    Learning to count anything: Reference-less class-agnostic counting with weak supervision.Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023

    Michael Hobley and Victor Prisacariu. Learning to count anything: Reference-less class-agnostic counting with weak supervision.Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023. 1

    2023

  8. [16]

    Drone- based object counting by spatially regularized regional pro- posal network

    Meng-Ru Hsieh, Yen-Liang Lin, and Winston H Hsu. Drone- based object counting by spatially regularized regional pro- posal network. InProceedings of the IEEE international conference on computer vision, pages 4145–4153, 2017. 5

    2017

  9. [17]

    Interac- tive class-agnostic object counting

    Yifeng Huang, Viresh Ranjan, and Minh Hoai. Interac- tive class-agnostic object counting. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 22312–22322, 2023. 1

    2023

  10. [18]

    Point segment and count: A gener- alized framework for object counting

    Zhizhong Huang, Mingliang Dai, Yi Zhang, Junping Zhang, and Hongming Shan. Point segment and count: A gener- alized framework for object counting. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 17067–17076, 2024. 6

    2024

  11. [19]

    Class- agnostic object counting with text-to-image diffusion model

    Xiaofei Hui, Qian Wu, Hossein Rahmani, and Jun Liu. Class- agnostic object counting with text-to-image diffusion model. InEuropean Conference on Computer Vision, pages 1–18. Springer, 2024. 1

    2024

  12. [20]

    Clip- count: Towards text-guided zero-shot object counting

    Ruixiang Jiang, Lingbo Liu, and Changwen Chen. Clip- count: Towards text-guided zero-shot object counting. In Proceedings of the 31st ACM International Conference on Multimedia, pages 4535–4545, 2023. 2, 6

    2023

  13. [21]

    Vlcounter: Text-aware visual representation for zero- shot object counting

    Seunggu Kang, WonJun Moon, Euiyeon Kim, and Jae-Pil Heo. Vlcounter: Text-aware visual representation for zero- shot object counting. InProceedings of the AAAI Conference on Artificial Intelligence, pages 2714–2722, 2024. 2, 6

    2024

  14. [22]

    Energy- based image deformation

    Zachi Karni, Daniel Freedman, and Craig Gotsman. Energy- based image deformation. InComputer Graphics Forum, pages 1257–1268. Wiley Online Library, 2009. 3

    2009

  15. [23]

    Segment any- thing

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer White- head, Alexander C Berg, Wan-Yen Lo, et al. Segment any- thing. InProceedings of the IEEE/CVF international confer- ence on computer vision, pages 4015–4026, 2023. 2, 6

    2023

  16. [24]

    Calibrating uncertainty for semi-supervised crowd counting

    Chen Li, Xiaoling Hu, Shahira Abousamra, and Chao Chen. Calibrating uncertainty for semi-supervised crowd counting. In2023 IEEE/CVF international conference on computer vision (ICCV), pages 16685–16695. IEEE, 2023. 1

    2023

  17. [25]

    Shuang Li, Faliang Chang, and Chunsheng Liu. Bi-directional dense traffic counting based on spatio-temporal counting fea- ture and counting-lstm network.IEEE Transactions on In- 9 telligent Transportation Systems, 22(12):7395–7407, 2020. 1

    2020

  18. [26]

    A simple-but-effective baseline for training-free class- agnostic counting

    Yuhao Lin, Haiming Xu, Lingqiao Liu, and Javen Qinfeng Shi. A simple-but-effective baseline for training-free class- agnostic counting. In2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pages 8155–8164. IEEE, 2025. 2

    2025

  19. [27]

    Improved baselines with visual instruction tuning

    Haotian Liu, Chunyuan Li, Yuheng Li, and Yong Jae Lee. Improved baselines with visual instruction tuning. InPro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 26296–26306, 2024. 5

    2024

  20. [28]

    Grounding dino: Marrying dino with grounded pre-training for open-set object detection

    Shilong Liu, Zhaoyang Zeng, Tianhe Ren, Feng Li, Hao Zhang, Jie Yang, Qing Jiang, Chunyuan Li, Jianwei Yang, Hang Su, et al. Grounding dino: Marrying dino with grounded pre-training for open-set object detection. InEuropean con- ference on computer vision, pages 38–55. Springe...

    2024

  21. [29]

    Countse: Soft exemplar open-set object counting

    Shuai Liu, Peng Zhang, Shiwei Zhang, and Wei Ke. Countse: Soft exemplar open-set object counting. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 21536–21546, 2025. 2

    2025

  22. [30]

    Can sam count anything? an empirical study on sam counting

    Zhiheng Ma, Xiaopeng Hong, and Qinnan Shangguan. Can sam count anything? an empirical study on sam counting. arXiv preprint arXiv:2304.10817, 2023. 2, 6

    2023

  23. [31]

    Through the magnifying glass: Adaptive perception magnification for hallucination-free vlm decoding

    Shunqi Mao, Chaoyi Zhang, and Weidong Cai. Through the magnifying glass: Adaptive perception magnification for hallucination-free vlm decoding. InProceedings of the 64th Annual Meeting of the Association for Computational Lin- guistics (Volume 1: Long Papers), pages 44480–4450...

    2026

  24. [32]

    Omnicount: Multi-label object counting with semantic- geometric priors

    Anindya Mondal, Sauradip Nag, Xiatian Zhu, and Anjan Dutta. Omnicount: Multi-label object counting with semantic- geometric priors. InProceedings of the AAAI Conference on Artificial Intelligence, pages 19537–19545, 2025. 2, 5, 6

    2025

  25. [33]

    Few-shot object counting and detection

    Thanh Nguyen, Chau Pham, Khoi Nguyen, and Minh Hoai. Few-shot object counting and detection. InEuropean Con- ference on Computer Vision, pages 348–365. Springer, 2022. 1

    2022

  26. [34]

    Sam3count for zero-shot open vocabulary counting in images and videos

    Joana Konadu Owusu and Shivanand Venkanna Sheshap- panavar. Sam3count for zero-shot open vocabulary counting in images and videos. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 581–591, 2026. 2, 6

    2026

  27. [35]

    Count- ingdino: A training-free pipeline for class-agnostic count- ing using unsupervised backbones

    Giacomo Pacini, Lorenzo Bianchi, Luca Ciampi, Nicola Messina, Giuseppe Amato, and Fabrizio Falchi. Count- ingdino: A training-free pipeline for class-agnostic count- ing using unsupervised backbones. InProceedings of the IEEE/CVF Winter Conference on Applications of Computer V...

    2026

  28. [36]

    Count-ception: Counting by fully convolutional redundant counting

    Joseph Paul Cohen, Genevieve Boucher, Craig A Glastonbury, Henry Z Lo, and Yoshua Bengio. Count-ception: Counting by fully convolutional redundant counting. InProceedings of the IEEE International conference on computer vision workshops, pages 18–26, 2017. 5

    2017

  29. [37]

    A novel unified architecture for low-shot counting by detection and segmentation.Advances in Neural Information Processing Systems, 37:66260–66282, 2024

    Jer Pelhan, Alan Lukezic, Vitjan Zavrtanik, and Matej Kris- tan. A novel unified architecture for low-shot counting by detection and segmentation.Advances in Neural Information Processing Systems, 37:66260–66282, 2024. 6

    2024

  30. [38]

    Dave-a detect-and-verify paradigm for low-shot counting

    Jer Pelhan, Vitjan Zavrtanik, Matej Kristan, et al. Dave-a detect-and-verify paradigm for low-shot counting. InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 23293–23302, 2024. 6

    2024

  31. [39]

    T2icount: Enhancing cross-modal understanding for zero-shot counting

    Yifei Qian, Zhongliang Guo, Bowen Deng, Chun Tong Lei, Shuai Zhao, Chun Pong Lau, Xiaopeng Hong, and Michael P Pound. T2icount: Enhancing cross-modal understanding for zero-shot counting. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 25336–25345,

  32. [40]

    Deep count: fruit counting based on deep simulated learning.Sensors, 17(4): 905, 2017

    Maryam Rahnemoonfar and Clay Sheppard. Deep count: fruit counting based on deep simulated learning.Sensors, 17(4): 905, 2017. 1

    2017

  33. [41]

    Learning to count everything

    Viresh Ranjan, Udbhav Sharma, Thu Nguyen, and Minh Hoai. Learning to count everything. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3394–3403, 2021. 2, 5

    2021

  34. [42]

    A comparative study of image retargeting

    Michael Rubinstein, Diego Gutierrez, Olga Sorkine, and Ariel Shamir. A comparative study of image retargeting. InACM SIGGRAPH Asia 2010 papers, pages 1–10. 2010. 3

    2010

  35. [43]

    Training-free ob- ject counting with prompts

    Zenglin Shi, Ying Sun, and Mengmi Zhang. Training-free ob- ject counting with prompts. InProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pages 323–331, 2024. 2, 6

    2024

  36. [44]

    Persense: Training-free personalized instance segmentation in dense images.BMVC, 2025

    Muhammad Ibraheem Siddiqui, Muhammad Umer Sheikh, Hassan Abid, and Muhammad Haris Khan. Persense: Training-free personalized instance segmentation in dense images.BMVC, 2025. 5

    2025

  37. [45]

    Learning to resize images for computer vision tasks

    Hossein Talebi and Peyman Milanfar. Learning to resize images for computer vision tasks. InProceedings of the IEEE/CVF international conference on computer vision, pages 497–506, 2021. 3

    2021

  38. [46]

    Degpr: Deep guided posterior regularization for multi-class cell de- tection and counting

    Aayush Kumar Tyagi, Chirag Mohapatra, Prasenjit Das, Govind Makharia, Lalita Mehra, Prathosh AP, et al. Degpr: Deep guided posterior regularization for multi-class cell de- tection and counting. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognit...

    2023

  39. [47]

    Zero-shot object counting

    Jingyi Xu, Hieu Le, Vu Nguyen, Viresh Ranjan, and Dim- itris Samaras. Zero-shot object counting. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 15548–15557, 2023. 2, 6

    2023

  40. [48]

    Class-agnostic few-shot object counting

    Shuo-Diao Yang, Hung-Ting Su, Winston H Hsu, and Wen- Chin Chen. Class-agnostic few-shot object counting. In Proceedings of the IEEE/CVF Winter Conference on Applica- tions of Computer Vision, pages 870–878, 2021. 1

    2021

  41. [49]

    Content-driven retargeting of stereoscopic images.IEEE Signal Processing Letters, 20(5):519–522, 2013

    Jin Woo Yoo, Sehoon Yea, and In Kyu Park. Content-driven retargeting of stereoscopic images.IEEE Signal Processing Letters, 20(5):519–522, 2013. 3

    2013

  42. [50]

    Few-shot object counting with similarity-aware feature enhancement

    Zhiyuan You, Kai Yang, Wenhan Luo, Xin Lu, Lei Cui, and Xinyi Le. Few-shot object counting with similarity-aware feature enhancement. InProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pages 6315– 6324, 2023. 1

    2023

  43. [51]

    Yolo-count: Differentiable object counting for text-to-image generation

    Guanning Zeng, Xiang Zhang, Zirui Wang, Haiyang Xu, Zeyuan Chen, Bingnan Li, and Zhuowen Tu. Yolo-count: Differentiable object counting for text-to-image generation. 10 InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 16765–16775, 2025. 6

    2025

  44. [52]

    Zero-shot object counting with vision-language prior guid- ance network.IEEE Transactions on Circuits and Systems for Video Technology, 2024

    Wenzhe Zhai, Xianglei Xing, Mingliang Gao, and Qilei Li. Zero-shot object counting with vision-language prior guid- ance network.IEEE Transactions on Circuits and Systems for Video Technology, 2024. 1

    2024

  45. [53]

    Boosting quantitive and spatial awareness for zero-shot object counting

    Da Zhang, Bingyu Li, Feiyu Wang, Zhiyuan Zhao, and Junyu Gao. Boosting quantitive and spatial awareness for zero-shot object counting. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20356– 20366, 2026. 2, 6

    2026

  46. [54]

    Personalize segment anything model with one shot

    Renrui Zhang, Zhengkai Jiang, Ziyu Guo, Shilin Yan, Junt- ing Pan, Hao Dong, Yu Qiao, Gao Peng, and Hongsheng Li. Personalize segment anything model with one shot. In International Conference on Learning Representations, pages 18250–18279, 2024. 2

    2024

  47. [55]

    Enhancing zero-shot object counting via text-guided local ranking and number- evoked global attention

    Shiwei Zhang, Qi Zhou, and Wei Ke. Enhancing zero-shot object counting via text-guided local ranking and number- evoked global attention. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 21097– 21106, 2025. 1

    2025

  48. [56]

    Instance-warp: Saliency guided image warping for unsuper- vised domain adaptation

    Shen Zheng, Anurag Ghosh, and Srinivasa Narasimhan. Instance-warp: Saliency guided image warping for unsuper- vised domain adaptation. InProceedings of the Winter Confer- ence on Applications of Computer Vision, pages 8186–8195,

  49. [57]

    apple” , GT:157 Count: 141 Spatial WarpingSimilarity Map Count: 159“stamp

    Huilin Zhu, Jingling Yuan, Zhengwei Yang, Yu Guo, Zheng Wang, Xian Zhong, and Shengfeng He. Zero-shot object counting with good exemplars. InEuropean Conference on Computer Vision, pages 368–385. Springer, 2024. 2, 6 11 A. Additional Qualitative Results Fig. 6 provides additio...

    2024

Pith tools

Reviewed July 3, 2026 · model on record in the stance chip above.