Pith. sign in

REVIEW 3 major objections 4 minor 51 references

Text-promptable Object Counting via Quantity Awareness Enhancement

T0 review · 3 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Counting by naming counts: quantity-aware text prompts lift text-promptable counting on FSC-147, CARPK, PUCPR+, and ShanghaiTech.

desk verdict Well-ablated counting paper with a real but fixable training-time leak confound in the prompt-feature ablation; worth reviewing, but the mechanism claim needs a clean control. read the letter →

arxiv 2507.06679 v1 pith:XMXPXC3B submitted 2025-07-09 cs.CV

Miaojing Shi , Xiaowen Zhang , Zijie Yue , Yong Luo , Cairong Zhao , Li Li This is my paper
classification cs.CV
keywords text-promptableobjectcountingquantityawarenessvision-languagemodeldensitymapestimationzero-shotcounterfactualtextpromptsdual-streamdecoderrankingloss
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 argues that text-promptable object counting fails to learn quantity because standard category prompts and contrastive losses treat all wrong numbers alike. To fix this, it proposes QUANet, which trains with quantity-oriented text prompts—'a photo of 14 kiwis' as a positive pair and 'a photo of 10/12/16/18 kiwis' as counterfactual negatives—and enforces an ordinal structure on image–text similarity scores. The claim is that this injected quantity awareness carries over to inference, when no number appears in the prompt, improving density-map counting. The paper also attributes gains to a dual-stream decoder that combines Transformer and CNN features through T2C-adapters and a cross-stream ranking loss. If right, it shows that auxiliary number-ordering supervision, not just category semantics, is what makes vision-language counters accurate.

What carries the argument

The central object is the quantity-oriented text prompt, written as 'a photo of [num] [class]' with the factual count $a_p$ and $N=7$ counterfactual counts $a_p \pm k\Delta$ for a dynamically chosen interval $\Delta$ proportional to the count magnitude. Around it sit three mechanisms: (1) the vision-text quantity alignment loss $L_{\text{align}}$ that orders image-text similarity scores by numerical distance from the true count; (2) the dual-stream adaptive counting decoder, which runs a Transformer stream and a CNN stream in parallel, with T2C-adapters (a cross-attention block plus a channel-excitation block) passing global Transformer knowledge into the CNN stream and a gating net fusing the two density maps; and (3) the cross-stream quantity ranking loss $L_{\text{rank}}$ that compares patch counts across and within streams with an interval $l=5$. The machine does its work by making quantity ordering a training signal, then converting the enhanced visual features into density predictions.

What would settle it

Train QUANet with the counterfactual counts' labels randomly permuted (so the ordinal structure of $L_{\text{align}}$ is broken while the positive/negative distinction stays); if test MAE on FSC-147 does not rise relative to the ordered version, the ordinal constraint is not the source of the gain. Alternatively, feed a fixed image with prompts 'a photo of k objects' for k = 1...20 and check whether predicted density changes monotonically; if it stays flat, the alignment did not transfer to inference.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a vision-language counter can be made quantity-aware by supervising the ordinal relation between an image and text prompts that state different object counts. QUANet constructs a factual prompt with the ground-truth count $a_p$ and counterfactual prompts with counts $a_p \pm k\Delta$, then applies a vision-text quantity alignment loss $L_{\text{align}}$ that forces the positive pair to have the highest cosine similarity and forces negative pairs to be ranked by numerical closeness to $a_p$. A cross-stream quantity ranking loss $L_{\text{rank}}$ additionally enforces patch-level count ordering inside and across the two decoder streams. With this training, the model reaches a test MAE of 13.24 on FSC-147, outperforming prior text-promptable methods (CountDiff: 14.83) despite removing the count from the prompt at inference, and generalizes zero-shot to CARPK (MAE 6.32), PUCPR+ (MAE 44.34), and ShanghaiTech SHA (MAE 140.2).

Load-bearing premise

The improvement depends on the model actually learning, from the auxiliary pairs, that a text number closer to the true count should look more similar to the image; if the vision–text embedding cannot represent that numerical closeness, the quantity alignment loss cannot be what improves counting.

Editorial extensions

If this is right

  • Text-promptable counting narrows the gap to exemplar-based counting: QUANet's 13.24 test MAE approaches CounTR (11.95) and LOCA (10.79) without needing manual exemplars.
  • The alignment loss's ordinal supervision transfers to inference: at test time the count is absent from the prompt, yet the model counts better, implying quantity awareness is baked into visual features rather than read off the text.
  • The DAC-decoder is a portable component: replacing the decoders of CLIP-Count and VLCounter with it improves their FSC-147 MAE, suggesting the dual-stream design generalizes beyond QUANet.
  • Adaptive intervals matter for counts spanning 7 to 1912: fixed intervals $\Delta=3,5,7,9$ all underperform the count-proportional $\Delta$ schedule.
  • Zero-shot domain transfer works: training only on FSC-147 yields strong results on drone car lots (CARPK) and dense crowds (ShanghaiTech) without fine-tuning.

Reading between the lines

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

  • The ordinal constraint between negative pairs is effectively a form of embedding-space regression: it teaches the model to represent 'how different' two numbers are, which could be extended to directly predicting counts from similarity slopes rather than from density integration.
  • If the mechanism is robust, synthetic counterfactual prompting could be applied to other VLM tasks where ordinal quantities matter, such as referring expression counting or visual grounding of 'more/less' relations.
  • A testable extension: train on FSC-147 with an expanded count range or synthesized images with known counts to see whether the alignment loss scales beyond the 1,000-object ceiling of the training data.
  • The direction of knowledge transfer (Transformer to CNN only) suggests that global contextual cues are most valuable when injected into local density estimation; a similar asymmetry may appear in other dense prediction tasks.
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

3 major / 4 minor

Summary. The paper introduces QUANet, a text-promptable object counting method that augments category prompts with explicit count information during training. A frozen BERT text encoder and a fine-tuned DINOv2 vision encoder are used with a novel quantity alignment loss applied to factual and counterfactual 'a photo of [num] [class]' prompts. The method also proposes a dual-stream adaptive counting decoder with Transformer and CNN streams, T2C-adapters, and a cross-stream quantity ranking loss. Experiments on FSC-147 and cross-dataset evaluations on CARPK, PUCPR+, and ShanghaiTech report state-of-the-art results among text-promptable methods, with extensive ablations of each component.

Significance. If the reported results and attributions hold, QUANet would be a meaningful advance in text-promptable class-agnostic counting, closing much of the gap to exemplar-based methods and demonstrating strong cross-dataset generalization. The paper's strengths include a clearly described loss formulation, systematic ablations of the proposed components, cross-dataset evaluation on four benchmarks, and public code release. However, the central mechanistic claim—that quantity-oriented prompts improve counting by enhancing quantity awareness in the vision encoder—is currently confounded by a training-time information leak in the category-conditioning feature, and the headline SOTA comparisons are partly confounded by a backbone change relative to prior methods.

major comments (3)
  1. [Section 3.2, Section 3.4, Table 4]
  2. [Section 4.2, Table 1]
  3. [Section 3.2, Eq. (2), Sec. 4.1]
minor comments (4)
  1. [Section 4.1]
  2. [Section 4.4, Fig. 5]
  3. [Section 3.4, Eq. (3)]
  4. [Table 5]

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the method is trained with ground-truth quantity prompts but evaluated on held-out test prompts that contain no count, and the reported gains are benchmarked externally.

full rationale

The paper's claimed derivation chain is not circular under the definitions used here. The quantity-oriented prompts (QTPs) are constructed from ground-truth counts during training only; at inference the paper explicitly replaces them with a category-only prompt ('a photo of birds'), so the reported test MAE values are genuine predictions and do not reduce to reading the count from the prompt. The counting loss, alignment loss, and ranking loss (Eqs. 1-4) are all supervised by ground-truth density/count on the FSC-147 training set, while the reported results are on held-out FSC-147 test, validation, and cross-dataset benchmarks (CARPK, PUCPR+, ShanghaiTech) without fine-tuning. The ablation tables (Tab. 4, Tab. 5, Tab. 8) compare variants using the same evaluation protocol, so they are standard empirical ablations rather than fitted inputs renamed as predictions. The skeptical concern about train-time leakage through the contextualized BERT embeddings (removing the [num] token does not fully erase count information from neighboring tokens) is a plausible attribution confound for interpreting the ablation gain, but it is not a definitional equivalence: no equation in the paper makes the test-time density output equal to the training-time count input. It is therefore a correctness/interpretability risk, not circularity. The paper also contains no load-bearing self-citation chain: the authors' prior works appear only in the related-work section on crowd counting and are not invoked as uniqueness theorems or as justification for the central loss or decoder design. The central components are evaluated against external baselines and held-out data, so the appropriate finding is no significant circularity.

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

No new physical entities are introduced. The method relies on standard pre-trained vision and text encoders, the assumption that embedding similarity can encode numerical ordering, and the reliability of density-map annotations for patch-level ranking.

free parameters (4)
  • Dynamic interval Delta mapping = {1,2,3,5,10,20,35,50}
    Hand-chosen scale steps to generate counterfactual counts per ground-truth count bin; these values determine the numerical distance of negative text prompts (Sec 4.1).
  • Number of counterfactual prompts N = 7
    Selected via validation (Fig 4a); formulas in Eq 2 appear to require an even N for the lower and upper halves, creating an inconsistency with the stated odd value.
  • Patch interval l in L_rank = 5
    Selected via validation (Fig 4b) to balance sensitivity of patch ranking.
  • Loss weight mu = 0.1
    Selected via validation (Fig 4c) for the combined alignment and ranking losses.
assumptions (4)
  • domain assumption Pre-trained DINOv2 (ViT-B/14) and BERT-base (top 9 blocks) provide a joint embedding space in which numerical proximity can be aligned.
    Sec 4.1 implementation details; the method relies on these frozen or lightly tuned encoders.
  • domain assumption Cosine similarity between global image and text embeddings is a monotone proxy for object count proximity.
    Central to L_align (Eq 2, Sec 3.4); if the embedding space cannot represent ordinal count differences, the auxiliary loss cannot inject quantity awareness.
  • domain assumption Ground truth density maps in FSC-147 support accurate patch-level rank supervision.
    Used by L_rank (Eq 3, Sec 3.4), which partitions density maps into 576 patches and ranks them by ground truth counts.
  • domain assumption Frozen text encoder preserves numerical semantics of number words.
    The BERT text encoder is frozen (Sec 4.1); L_align requires '14 kiwis' to be closer to a 14-object image than '12 kiwis' is.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Text-promptable Object Counting via Quantity Awareness Enhancement." pith.science (2026). https://pith.science/paper/XMXPXC3B

@misc{pith2026250706679,
  author       = {Pith},
  title        = {Pith review of: Text-promptable Object Counting via Quantity Awareness Enhancement},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XMXPXC3B}},
  note         = {Machine review of arXiv:2507.06679}
}
read the original abstract

Recent advances in large vision-language models (VLMs) have shown remarkable progress in solving the text-promptable object counting problem. Representative methods typically specify text prompts with object category information in images. This however is insufficient for training the model to accurately distinguish the number of objects in the counting task. To this end, we propose QUANet, which introduces novel quantity-oriented text prompts with a vision-text quantity alignment loss to enhance the model's quantity awareness. Moreover, we propose a dual-stream adaptive counting decoder consisting of a Transformer stream, a CNN stream, and a number of Transformer-to-CNN enhancement adapters (T2C-adapters) for density map prediction. The T2C-adapters facilitate the effective knowledge communication and aggregation between the Transformer and CNN streams. A cross-stream quantity ranking loss is proposed in the end to optimize the ranking orders of predictions from the two streams. Extensive experiments on standard benchmarks such as FSC-147, CARPK, PUCPR+, and ShanghaiTech demonstrate our model's strong generalizability for zero-shot class-agnostic counting. Code is available at https://github.com/viscom-tongji/QUANet

Figures

Figures reproduced from arXiv: 2507.06679 by the authors.

Figure 1
Figure 1. (a) Previous text-promptable object counting methods [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. (a) Overall training architecture of our QUANet. Given a query image, we first craft the quantity-oriented text prompts to specify [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Visual comparison between different text-promptable object counting methods. Counting values are shown at the right top of the [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Parameter variation analysis on (a) the number [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 6
Figure 6. Figure 6: Sample sequences with ranked predictions. [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 5
Figure 5. Figure 5: Similarities between positive vision-text pairs. [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

51 extracted references · 44 canonical work pages

  1. [1]

    Amini-Naieni, K

    N. Amini-Naieni, K. Amini-Naieni, T. Han, and A. Zisser- man. Open-world text-specified object counting. In British Machine Vision Conference, 2023. 2, 3, 4, 6, 7, 9, 10

    work page 2023

  2. [2]

    Counting in the wild

    Carlos Arteta, Victor Lempitsky, and Andrew Zisserman. Counting in the wild. In Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, Octo- ber 11–14, 2016, Proceedings, Part VII 14 , pages 483–498. Springer, 2016. 2

    work page 2016

  3. [3]

    Single domain generalization for few-shot counting via universal representation matching

    Xianing Chen, Si Huo, Borui Jiang, Hailin Hu, and Xinghao Chen. Single domain generalization for few-shot counting via universal representation matching. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 4639–4649, 2025. 2

    work page 2025

  4. [4]

    Reproducible scal- ing laws for contrastive language-image learning

    Mehdi Cherti, Romain Beaumont, Ross Wightman, Mitchell Wortsman, Gabriel Ilharco, Cade Gordon, Christoph Schuh- mann, Ludwig Schmidt, and Jenia Jitsev. Reproducible scal- ing laws for contrastive language-image learning. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2818–2829, 2023. 3

    work page 2023

  5. [5]

    Referring ex- pression counting

    Siyang Dai, Jun Liu, and Ngai-Man Cheung. Referring ex- pression counting. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 16985–16995, 2024. 2, 3

    work page 2024

  6. [6]

    Bert: Pre-training of deep bidirectional transformers for language understanding

    Jacob Devlin. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805, 2018. 6

    arXiv 2018

  7. [7]

    An image is worth 16x16 words: Trans- formers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, et al. An image is worth 16x16 words: Trans- formers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020. 2

    arXiv 2010

  8. [8]

    Semantic generative augmentations for few-shot counting

    Perla Doubinsky, Nicolas Audebert, Michel Crucianu, and Hervé Le Borgne. Semantic generative augmentations for few-shot counting. In Proceedings of the IEEE/CVF Win- ter Conference on Applications of Computer Vision , pages 5443–5452, 2024. 2

    work page 2024

Show all 51 references
  1. [9]

    Domain- general crowd counting in unseen scenarios

    Zhipeng Du, Jiankang Deng, and Miaojing Shi. Domain- general crowd counting in unseen scenarios. In Proceedings of the AAAI Conference on Artificial Intelligence, pages 561– 570, 2023. 2

    2023

  2. [10]

    Regressor-segmenter mutual prompt learning for crowd counting

    Mingyue Guo, Li Yuan, Zhaoyi Yan, Binghui Chen, Yaowei Wang, and Qixiang Ye. Regressor-segmenter mutual prompt learning for crowd counting. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 28380–28389, 2024. 2

    2024

  3. [11]

    Learning to count anything: Reference-less class-agnostic counting with weak supervision

    Michael Hobley and Victor Prisacariu. Learning to count anything: Reference-less class-agnostic counting with weak supervision. Proceedings of the IEEE Conference on Com- puter Vision and Pattern Recognition (CVPR), 2023. 7

    2023

  4. [12]

    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. In Proceedings of the IEEE international conference on computer vision, pages 4145–4153, 2017. 2, 6

    2017

  5. [13]

    Squeeze-and-excitation net- works

    Jie Hu, Li Shen, and Gang Sun. Squeeze-and-excitation net- works. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 7132–7141, 2018. 4

    2018

  6. [14]

    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. In European Conference on Computer Vision, pages 1–18. Springer, 2024. 6, 7

    2024

  7. [15]

    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. arXiv preprint arXiv:2305.07304, 2023. 2, 3, 4, 6, 7, 9

    2023 arXiv

  8. [16]

    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. In Proceedings of the AAAI Conference on Artificial Intelligence, pages 2714–2722, 2024. 2, 3, 6, 7, 9, 10

    2024

  9. [17]

    Calibrating uncertainty for semi-supervised crowd counting

    Chen Li, Xiaoling Hu, Shahira Abousamra, and Chao Chen. Calibrating uncertainty for semi-supervised crowd counting. In 2023 IEEE/CVF International Conference on Computer Vision (ICCV), pages 16685–16695. IEEE, 2023. 2

    2023

  10. [18]

    Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models

    Junnan Li, Dongxu Li, Silvio Savarese, and Steven Hoi. Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models. In In- ternational conference on machine learning , pages 19730– 19742. PMLR, 2023. 1

    2023

  11. [19]

    Csrnet: Di- lated convolutional neural networks for understanding the highly congested scenes

    Yuhong Li, Xiaofan Zhang, and Deming Chen. Csrnet: Di- lated convolutional neural networks for understanding the highly congested scenes. In Proceedings of the IEEE con- ference on computer vision and pattern recognition , pages 1091–1100, 2018. 8

    2018

  12. [20]

    An end-to-end transformer model for crowd localization

    Dingkang Liang, Wei Xu, and Xiang Bai. An end-to-end transformer model for crowd localization. European Confer- ence on Computer Vision, 2022. 1, 2

    2022

  13. [21]

    Crowdclip: Unsupervised crowd counting via vision-language model

    Dingkang Liang, Jiahao Xie, Zhikang Zou, Xiaoqing Ye, Wei Xu, and Xiang Bai. Crowdclip: Unsupervised crowd counting via vision-language model. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 2893–2903, 2023. 2 11

    2023

  14. [22]

    A fixed-point approach to unified prompt-based counting

    Wei Lin and Antoni B Chan. A fixed-point approach to unified prompt-based counting. In Proceedings of the AAAI Conference on Artificial Intelligence , pages 3468– 3476, 2024. 6

    2024

  15. [23]

    Countr: Transformer-based generalised visual count- ing

    Chang Liu, Yujie Zhong, Andrew Zisserman, and Weidi Xie. Countr: Transformer-based generalised visual count- ing. arXiv preprint arXiv:2208.13721, 2022. 2, 3, 4, 6, 7

    2022 arXiv

  16. [24]

    Point-query quadtree for crowd counting, localization, and more

    Chengxin Liu, Hao Lu, Zhiguo Cao, and Tongliang Liu. Point-query quadtree for crowd counting, localization, and more. In Proceedings of the IEEE/CVF International Con- ference on Computer Vision, pages 1676–1685, 2023. 2, 6

    2023

  17. [25]

    Visual instruction tuning

    Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. Visual instruction tuning. Advances in neural information processing systems, 36, 2024. 1

    2024

  18. [26]

    Point in, box out: Beyond counting persons in crowds

    Yuting Liu, Miaojing Shi, Qijun Zhao, and Xiaofang Wang. Point in, box out: Beyond counting persons in crowds. In Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 6469–6478, 2019. 2

    2019

  19. [27]

    Discovering regression- detection bi-knowledge transfer for unsupervised cross- domain crowd counting

    Yuting Liu, Zheng Wang, Miaojing Shi, Shin’ichi Satoh, Qijun Zhao, and Hongyu Yang. Discovering regression- detection bi-knowledge transfer for unsupervised cross- domain crowd counting. Neurocomputing, 494:418–431,

  20. [28]

    Decoupled weight decay regularization

    I Loshchilov. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101, 2017. 6

    2017 arXiv

  21. [29]

    Class-agnostic counting

    Erika Lu, Weidi Xie, and Andrew Zisserman. Class-agnostic counting. In Computer Vision–ACCV 2018: 14th Asian Con- ference on Computer Vision, Perth, Australia, December 2– 6, 2018, Revised Selected Papers, Part III 14 , pages 669–

    2018

  22. [30]

    A large contextual dataset for classification, detection and counting of cars with deep learning

    T Nathan Mundhenk, Goran Konjevod, Wesam A Sakla, and Kofi Boakye. A large contextual dataset for classification, detection and counting of cars with deep learning. In Com- puter Vision–ECCV 2016: 14th European Conference, Am- sterdam, The Netherlands, October 11-14, 2016, Pro...

    2016

  23. [31]

    Dinov2: Learning robust visual features without supervision

    Maxime Oquab, Timothée Darcet, Théo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual features without supervision. arXiv preprint arXiv:2304.07193, 2023. 6

    2023 arXiv

  24. [32]

    Teaching clip to count to ten

    Roni Paiss, Ariel Ephrat, Omer Tov, Shiran Zada, Inbar Mosseri, Michal Irani, and Tali Dekel. Teaching clip to count to ten. In Proceedings of the IEEE/CVF International Con- ference on Computer Vision, pages 3170–3180, 2023. 2, 3, 9

    2023

  25. [33]

    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. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 23293–23302, 2024. 2, 6, 9, 10

    2024

  26. [34]

    Single domain general- ization for crowd counting

    Zhuoxuan Peng and S-H Gary Chan. Single domain general- ization for crowd counting. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 28025–28034, 2024. 2, 6

    2024

  27. [35]

    Learning transferable visual models from natural language supervi- sion

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervi- sion. In International conference on machine learning, ...

    2021

  28. [36]

    Crowd- diff: Multi-hypothesis crowd density estimation using dif- fusion models

    Yasiru Ranasinghe, Nithin Gopalakrishnan Nair, Wele Gedara Chaminda Bandara, and Vishal M Patel. Crowd- diff: Multi-hypothesis crowd density estimation using dif- fusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 12809– ...

    2024

  29. [37]

    Exemplar free class agnostic counting

    Viresh Ranjan and Minh Hoai Nguyen. Exemplar free class agnostic counting. In Proceedings of the Asian Conference on Computer Vision, pages 3121–3137, 2022. 1, 3

    2022

  30. [38]

    Learning to count everything

    Viresh Ranjan, Udbhav Sharma, Thu Nguyen, and Minh Hoai. Learning to count everything. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3394–3403, 2021. 2, 6, 7

    2021

  31. [39]

    Re- visiting perspective information for efficient crowd counting

    Miaojing Shi, Zhaohui Yang, Chao Xu, and Qijun Chen. Re- visiting perspective information for efficient crowd counting. In Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 7279–7288, 2019. 2

    2019

  32. [40]

    Represent, compare, and learn: A similarity-aware framework for class-agnostic counting

    Min Shi, Hao Lu, Chen Feng, Chengxin Liu, and Zhiguo Cao. Represent, compare, and learn: A similarity-aware framework for class-agnostic counting. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 9529–9538, 2022. 2, 6, 7

    2022

  33. [41]

    A low-shot object counting network with iterative prototype adaptation

    Nikola Ðuki ´c, Alan Lukeži ˇc, Vitjan Zavrtanik, and Matej Kristan. A low-shot object counting network with iterative prototype adaptation. In Proceedings of the IEEE/CVF In- ternational Conference on Computer Vision , pages 18872– 18881, 2023. 2, 6

    2023

  34. [42]

    Exploring contextual at- tribute density in referring expression counting

    Zhicheng Wang, Zhiyu Pan, Zhan Peng, Jian Cheng, Liwen Xiao, Wei Jiang, and Zhiguo Cao. Exploring contextual at- tribute density in referring expression counting. In Proceed- ings of the Computer Vision and Pattern Recognition Con- ference, pages 19587–19596, 2025. 3

    2025

  35. [43]

    Learning super-features for image re- trieval

    Philippe Weinzaepfel, Thomas Lucas, Diane Larlus, and Yannis Kalantidis. Learning super-features for image re- trieval. In ICLR, 2022. 10

    2022

  36. [44]

    Boosting detection in crowd analysis via underutilized output features

    Shaokai Wu and Fengyu Yang. Boosting detection in crowd analysis via underutilized output features. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 15609–15618, 2023. 2

    2023

  37. [45]

    Mi- croscopy cell counting and detection with fully convolutional regression networks

    Weidi Xie, J Alison Noble, and Andrew Zisserman. Mi- croscopy cell counting and detection with fully convolutional regression networks. Computer methods in biomechanics and biomedical engineering: Imaging & Visualization, 6(3): 283–292, 2018. 1, 2

    2018

  38. [46]

    Zero-shot object counting

    Jingyi Xu, Hieu Le, Vu Nguyen, Viresh Ranjan, and Dim- itris Samaras. Zero-shot object counting. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 15548–15557, 2023. 3, 5, 6, 7

    2023

  39. [47]

    Pbe- count: Prompt-before-extract paradigm for class-agnostic counting

    Canchen Yang, Tianyu Geng, Jian Peng, and Chun Xu. Pbe- count: Prompt-before-extract paradigm for class-agnostic counting. In Proceedings of the AAAI Conference on Arti- ficial Intelligence, pages 9139–9147, 2025. 2 12

    2025

  40. [48]

    Zero-shot object counting with vision-language prior guid- ance network

    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, 35(3):2487–2498, 2025. 2, 3, 6, 7

    2025

  41. [49]

    Single-image crowd counting via multi-column convolutional neural network

    Yingying Zhang, Desen Zhou, Siqin Chen, Shenghua Gao, and Yi Ma. Single-image crowd counting via multi-column convolutional neural network. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 589–597, 2016. 2, 6

    2016

  42. [50]

    Zero-shot object counting with good exemplars

    Huilin Zhu, Jingling Yuan, Zhengwei Yang, Yu Guo, Zheng Wang, Xian Zhong, and Shengfeng He. Zero-shot object counting with good exemplars. In European Conference on Computer Vision, pages 368–385. Springer, 2024. 3, 6, 7, 9, 10 13

    2024

  43. [684]

    Springer, 2019. 1, 2, 6

    2019

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.