Pith. sign in

REVIEW 3 major objections 6 minor 43 references

Attend To Count: Crowd Counting with Adaptive Capacity Multi-scale CNNs

T0 review · 3 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read ACM-CNN assigns different network capacity to dense and sparse crowd regions, cutting count error on five benchmarks.

desk verdict Incrementally novel attention-based crowd counter with a plausible design, but the headline results rest on test-set tuning and an unvalidated assumption that coarse peaks mark dense regions. read the letter →

arxiv 1908.02797 v2 pith:QNTJFF4I submitted 2019-08-07 cs.CV

Zhikang Zou , Yu Cheng , Xiaoye Qu , Shouling Ji , Xiaoxiao Guo , Pan Zhou This is my paper · ORCID
classification cs.CV
keywords crowdcountingdensitymapestimationcountattentionadaptivecapacitymulti-scaleCNNcoarse-to-finenetworksurveillanceanalysismechanism
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 tries to show that crowd counting improves when a model does not process every image region with the same fixed network. It proposes a three-part architecture: a coarse network first produces a rough density map, a count attention mechanism selects the most densely populated patches from that map, and a fine network processes only those patches; a smooth network then fuses the coarse and fine outputs into the final density map. On five public datasets the authors report lower mean absolute error than the baselines they compare against, on both dense and sparse scenes. The point of the work is that capacity allocation, not just scale diversity, is what a counting network should learn.

What carries the argument

The load-bearing object is the count attention mechanism. It traverses the coarse density map $C_0$, selects the $N$ positions with the highest pixel values as centers $\text{pos}_i$, crops the corresponding spatial patches $\text{CSM}_i$ from the input, and accumulates them into the high-density map $S_{N+1} = S_1 + \sum_i \text{CSM}_i \otimes S_0$, while the complement $C_{N+1} = C_1 - \sum_i \text{CSM}_i \otimes C_0$ keeps the low-density regions. This pair of operations is what lets one network spend deep capacity exactly where the coarse estimate says people are dense. The secondary machinery is the smooth network, a small stack of convolution-ReLU layers that fuses the fine feature map with the low-density coarse map so the final density map does not show seams at the attention patch boundaries.

What would settle it

On a held-out set of crowd images, compare the top-$N$ patch centers picked by the count attention mechanism against the actual positions of the highest-density head clusters in the ground-truth density map. If the selected centers miss a substantial fraction of true dense clusters while the final MAE stays low, the attention mechanism is not the cause of the accuracy; conversely, if errors concentrate in missed clusters, the premise is confirmed.

Watch

Extended reading notes

Core claim

The central claim is that adaptive capacity allocation beats a single fixed-capacity structure for crowd counting: a shallow multi-column network is good at finding where crowds are, a deep network is good at resolving dense regions, and combining them through an attention-guided split yields accurate density maps. The count attention mechanism reads the coarse density map, takes the highest-valued pixel positions as attention centers, crops patches around them, and iteratively builds a high-density input $S_{N+1}$ for the fine network while stripping those same regions from the coarse feature map to form the low-density map $C_{N+1}$. The fine and coarse features are added, then passed through a small smoothing network to remove the rigid boundary between attended and unattended regions. The authors report MAE reductions to 72.2 on ShanghaiTech Part A, 1.01 on UCSD, and 291.6 on UCF CC 50, and ablation results attributing the gain to the attention mechanism and the smoothing fusion.

Load-bearing premise

The mechanism assumes the highest-valued pixels in the coarse density map are the true dense crowd centers, so if the coarse network peaks in the wrong places, the fine network refines irrelevant patches and the claimed benefit of adaptive capacity does not follow.

Editorial extensions

If this is right

  • On dense-scene benchmarks, the full three-network model lowers MAE from 74.5 to 72.2 on ShanghaiTech Part A when the smooth network is added, so the smoothing stage contributes accuracy, not just visual smoothness.
  • On sparse-scene benchmarks, the same architecture reaches MAE 1.01 on UCSD and 2.3 on Mall, showing the adaptive split does not hurt when crowds occupy only a small part of the frame.
  • The count attention mechanism is not tied to the specific coarse and fine networks: swapping in the deep and shallow networks from CrowdNet still improves UCSD MAE from 1.5 with simple concatenation to 1.2 with count attention, indicating it is a general fusion strategy.
  • Because the method is fully convolutional and trained end-to-end with an $L_2$ loss plus an intermediate supervision term, it can be inserted into existing counting pipelines without changing the training paradigm.

Reading between the lines

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

  • A natural extension is to make the attention selection differentiable or soft instead of a hard top-$N$ crop; this could let the model learn how many patches and what sizes are needed per image and remove the dataset-dependent threshold parameter.
  • If the coarse network's peaks are unreliable, the error propagates: a missed dense cluster never reaches the fine network. A testable prediction of the paper's mechanism is that counting error concentrates in clusters the coarse map ranks below the top $N$.
  • The same attend-then-refine loop could transfer to other dense-object counting tasks, such as cell counting in microscopy or vehicle counting in aerial imagery, where a cheap density estimate can gate a deeper refinement network.
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

3 major / 6 minor

Summary. This paper proposes ACM-CNN, a crowd counting architecture with three modules: a shallow multi-column coarse network, a VGG-16 based fine network, and a smooth fusion network. A 'count attention' mechanism selects the N highest-valued pixels in the coarse density map C0, extracts the corresponding patches from the input for the fine network, and strips those patches from C0 before the two feature maps are summed and passed to the smooth network. The authors report state-of-the-art or near-state-of-the-art MAE/MSE on ShanghaiTech Part A/B, WorldExpo'10, UCSD, Mall, and UCF CC 50, and provide ablations showing each module contributes, plus a generality experiment using CrowdNet sub-networks.

Significance. The adaptive-capacity idea is reasonable and the three-module decomposition is clearly described. The paper includes useful ablations (Table 8) and a generality experiment (Table 9). If the reported numbers are reproducible, the method would be competitive. However, the significance is currently limited by evaluation concerns: hyperparameters N and S appear to be selected on test sets, no error bars are reported, and no code is released, so the state-of-the-art claim rests on weaker evidence than the tables suggest.

major comments (3)
  1. [Section 3.1-3.2, Eqs. (5)-(6), (8)] The count attention mechanism assumes that the highest-valued pixels in the coarse map C0 identify dense crowd regions, but this property is never supervised or validated. The intermediate loss Lb in Eq. (8) supervises the merged map Mb = CN+1 + F, and Eq. (5) strips the selected C0 regions out before this sum; the hard top-N selection is non-differentiable, so no gradient encourages C0 to place its peaks in truly dense areas. The manuscript provides no visualization or quantitative check that the selected patches coincide with dense regions (e.g., an IoU measure against the ground-truth density or a comparison of the selected patches with the true head locations). Without such evidence, the improvement of C+F over C or F in Table 8 cannot be attributed to correct localization rather than simply to the added capacity of a second network.
  2. [Section 5.1, Tables 1-5] The hyperparameters N and S appear to be selected on the test sets. The dataset splits described in Sections 4.2 and 4.4 contain no validation portion, and Figure 9 plots MAE against N and S for Shanghaitech Part A and UCSD without distinguishing train, validation, or test. Choosing N=1, S=[0.9*height,0.9*width] for dense sets and N=5, S=[70,70] for sparse sets from these curves and then reporting the resulting test numbers as state-of-the-art risks test-set overfitting. The authors should tune on a held-out validation split (or fix N and S before evaluation) and report the mean and standard deviation over multiple runs, especially for the 5-fold cross-validation on UCF CC 50 in Table 5.
  3. [Section 5.1, Table 7] For all dense datasets, the chosen configuration is N=1 with a patch covering 0.9*height by 0.9*width, i.e., 81% of the image area. Under this configuration, the fine network processes nearly the entire image and the coarse network contributes only the remaining border regions; this weakens the paper's central claim of assigning 'different capacities to different portions' adaptively. The authors should either justify this configuration as a limiting case of adaptive capacity or provide results for smaller patches to demonstrate that spatial selection is what drives the gain.
minor comments (6)
  1. [Section 2.2] There is a typo: 'convolutional nerual network' should be 'convolutional neural network'.
  2. [Section 3.3] There is a typo: 'the mooth network' should be 'the smooth network'.
  3. [Table 9] The label 'CrowdNet contact' should be 'CrowdNet concatenation'.
  4. [Section 4.1, Eqs. (11)-(12)] The notation in the MAE and MSE equations uses a norm symbol where an absolute value is intended; please replace with |ci - ci_hat| for clarity.
  5. [Table 2] The method 'LBP + RR' is cited as reference [5], but [5] is MCNN; please correct the citation.
  6. [Figure 2] The architecture diagram is very dense; the count attention submodule, in particular, would benefit from a separate, more detailed illustration.

Circularity Check

0 steps flagged · score 0.0 of 10

No derivation-chain circularity: the final density map is supervised end-to-end and the count-attention routing is not a restatement of the coarse map; test-set hyperparameter selection is an evaluation concern, not a circular reduction.

full rationale

ACM-CNN is an empirical architecture paper, not a derivation. The final density map Ma is obtained by merging the fine-network output F with the residual low-density map CN+1 and passing through a smooth network (Eqs. 5-7, 9-10), and all components are trained against the same ground-truth density map D via Lb and La. The count-attention selection (Eqs. 3-4) uses the coarse prediction C0 to choose crop locations, but the final output is not defined to equal C0: the selected regions are re-processed by the fine network and the coarse contribution is explicitly removed (Eq. 5), so there is no equation-level identity that makes the prediction a restatement of its inputs. The 'highest values indicate dense areas' statement is a property of the constructed ground truth (Eq. 2) and is applied to the predicted C0 as an empirical localization assumption; this is an unsupported premise and a correctness risk, not a circular definition. The paper does not rely on any load-bearing self-citation: prior-work citations (MCNN, VGG, CrowdNet, etc.) are external baselines or architecture choices, and no uniqueness theorem or fitted parameter is invoked to force the result. The only notable concern is that N and S are tuned using the test results in Section 5.1 and Table 7, but this is benchmark leakage under soundness, not a circularity of the model's input-output relation. Therefore no circular step is exhibited.

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

The central claim rests on the density map formation rule, the reliability of coarse predictions as attention guides, and the hand-chosen hyperparameters N, S, T, beta, and k. No new physical or conceptual entities are introduced; the contribution is a specific network architecture and routing mechanism.

free parameters (4)
  • N (number of attention regions) = N=1 for ShanghaiTech, WorldExpo'10, UCF CC 50; N=5 for UCSD, Mall
    Chosen by grid search on ShanghaiTech Part A and UCSD test sets (Section 5.1, Figure 9); applied to other datasets by dense/sparse classification.
  • S (attention patch size) = [0.9*height, 0.9*width] for dense datasets; [0.4*height, 0.3*width] for sparse datasets
    Selected together with N by test-set experiments (Section 5.1, Figure 9, Table 7).
  • T (dense/sparse threshold) = 40 average count
    Hand-chosen to classify datasets into dense and sparse (Section 5.1, Table 6); affects which N and S values are used.
  • beta and k in density map generation = beta=0.3, k=3
    Follows MCNN and controls Gaussian kernel widths for ground truth density maps (Section 3.1, Eq. 2).
assumptions (4)
  • domain assumption Geometry-adaptive Gaussian kernels produce a valid regression target for crowd counting.
    Invoked in Section 3.1, Eq. 2 to build ground truth density maps; taken from MCNN without re-validation.
  • domain assumption The highest pixel values in the coarse density map indicate the most densely populated image regions.
    Core premise of count attention (Section 3.1, sentence after Eq. 2); if false, the fine network refines the wrong regions.
  • domain assumption Deep networks perform better on dense regions and shallow networks on sparse regions.
    Motivates the coarse/fine split (Introduction and Section 5.2); the paper provides ablations on two datasets but it remains a modeling assumption.
  • standard math L2 loss between predicted and ground truth density maps is an appropriate training objective.
    Used in Eqs. 8-10; standard for density map regression.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Attend To Count: Crowd Counting with Adaptive Capacity Multi-scale CNNs." pith.science (2026). https://pith.science/paper/QNTJFF4I

@misc{pith2026190802797,
  author       = {Pith},
  title        = {Pith review of: Attend To Count: Crowd Counting with Adaptive Capacity Multi-scale CNNs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QNTJFF4I}},
  note         = {Machine review of arXiv:1908.02797}
}
read the original abstract

Crowd counting is a challenging task due to the large variations in crowd distributions. Previous methods tend to tackle the whole image with a single fixed structure, which is unable to handle diverse complicated scenes with different crowd densities. Hence, we propose the Adaptive Capacity Multi-scale convolutional neural networks (ACM-CNN), a novel crowd counting approach which can assign different capacities to different portions of the input. The intuition is that the model should focus on important regions of the input image and optimize its capacity allocation conditioning on the crowd intensive degree. ACM-CNN consists of three types of modules: a coarse network, a fine network, and a smooth network. The coarse network is used to explore the areas that need to be focused via count attention mechanism, and generate a rough feature map. Then the fine network processes the areas of interest into a fine feature map. To alleviate the sense of division caused by fusion, the smooth network is designed to combine two feature maps organically to produce high-quality density maps. Extensive experiments are conducted on five mainstream datasets. The results demonstrate the effectiveness of the proposed model for both density estimation and crowd counting tasks.

Figures

Figures reproduced from arXiv: 1908.02797 by the authors.

Figure 1
Figure 1. Top row indicates typical static crowd scenes from the [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Architecture of the proposed Adaptive Capacity Multi-scale convolutional neural networks. The coarse network takes the input [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The use of this smooth network contributes to [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Visualization of the crowd density maps by various meth [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Example results on WorldExpo’10 dataset. Each row rep [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Results on UCSD dataset using the proposed ACM-CNN. [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: Example results on Mall dataset using our ACM-CNN. [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 9
Figure 9. Figure 9: Histograms of two datasets on parameter selection. The entire chart is built according to the changes in the two parameters: the [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]
Figure 10
Figure 10. Figure 10: Predictions of various structures introduced in [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 30 canonical work pages

  1. [1]

    Viola, M

    P. Viola, M. J. Jones, D. Snow, Detecting pedestrians using patterns of motion and appearance, in: null, IEEE, 2003, p. 734

    work page 2003

  2. [2]

    Y. P. Kocak, S. Sevgen, Detecting and counting people using real-time directional algorithms implemented by compute uni- fied device architecture, Neurocomputing 248 (2017) 105 – 111, neural Networks : Learning Algorithms and Classification Sys- tems

    work page 2017

  3. [3]

    A. B. Chan, N. Vasconcelos, Bayesian poisson regression for crowd counting, in: Computer Vision, 2009 IEEE 12th Interna- tional Conference on, IEEE, 2009, pp. 545–551

    work page 2009

  4. [4]

    K. Chen, C. C. Loy, S. Gong, T. Xiang, Feature mining for localised crowd counting., in: BMVC, Vol. 1, 2012, p. 3

    work page 2012

  5. [5]

    Zhang, D

    Y. Zhang, D. Zhou, S. Chen, S. Gao, Y. Ma, Single-image crowd counting via multi-column convolutional neural network, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 589–597

    work page 2016

  6. [6]

    Crowd counting via scale-adaptive convolutional neural network

    L. Zhang, M. Shi, Q. Chen, Crowd counting via scale-adaptive convolutional neural network, CoRR abs/1711.04433

    work page Pith review arXiv

  7. [7]

    D. B. Sam, S. Surya, R. V. Babu, Switching convolutional neural network for crowd counting, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Vol. 1, 2017, p. 6

    work page 2017

  8. [8]

    Simonyan, A

    K. Simonyan, A. Zisserman, Very deep convolutional networks for large-scale image recognition, CoRR abs/1409.1556

    arXiv

Show all 43 references
  1. [9]

    Q. Wang, J. Gao, W. Lin, Y. Yuan, Learning from synthetic data for crowd counting in the wild, CoRR abs/1903.03303

    1903 arXiv

  2. [10]

    Q. Wang, M. Chen, F. Nie, X. Li, Detecting coherent groups in crowd scenes by multiview clustering, IEEE Transactions on Pattern Analysis and Machine Intelligence (2018) 1–1 doi:10. 1109/TPAMI.2018.2875002

    2018

  3. [11]

    Q. Wang, J. Wan, Y. Yuan, Deep metric learning for crowd- edness regression, IEEE Transactions on Circuits and Sys- tems for Video Technology 28 (10) (2018) 2633–2643. doi: 10.1109/TCSVT.2017.2703920. 9

    2018

  4. [12]

    Zhang, C

    Y. Zhang, C. Zhou, F. Chang, A. C. Kot, Multi-resolution at- tention convolutional neural network for crowd counting, Neu- rocomputing 329 (2019) 144 – 152. doi:https://doi.org/10. 1016/j.neucom.2018.10.058

    2019

  5. [13]

    V. A. Sindagi, V. M. Patel, A survey of recent advances in cnn- based single image crowd counting and density estimation, Pat- tern Recognition Letters 107 (2018) 3 – 16, video Surveillance- oriented Biometrics. doi:https://doi.org/10.1016/j.patrec. 2017.07.007

    2018 doi

  6. [14]

    Jiang, Z

    X. Jiang, Z. Xiao, B. Zhang, X. Zhen, X. Cao, D. S. Doer- mann, L. Shao, Crowd counting and density estimation by trel- lis encoder-decoder network, CoRR abs/1903.00853

    1903 arXiv

  7. [15]

    N. Liu, Y. Long, C. Zou, Q. Niu, L. Pan, H. Wu, Adcrowd- net: An attention-injective deformable convolutional network for crowd understanding, CoRR abs/1811.11968

    arXiv

  8. [16]

    Y. Liu, M. Shi, Q. Zhao, X. Wang, Point in, box out: Beyond counting persons in crowds, CoRR abs/1904.01333

    1904 arXiv

  9. [17]

    V. A. Sindagi, V. M. Patel, Generating high-quality crowd den- sity maps using contextual pyramid cnns, in: The IEEE Inter- national Conference on Computer Vision (ICCV), 2017

    2017

  10. [18]

    Babu Sam, N

    D. Babu Sam, N. N. Sajjan, R. Venkatesh Babu, M. Srinivasan, Divide and grow: Capturing huge diversity in crowd images with incrementally growing cnn, in: The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018

    2018

  11. [19]

    X. Cao, Z. Wang, Y. Zhao, F. Su, Scale aggregation network for accurate and efficient crowd counting, in: The European Conference on Computer Vision (ECCV), 2018

    2018

  12. [20]

    Y. Li, X. Zhang, D. Chen, Csrnet: Dilated convolutional neural networks for understanding the highly congested scenes, in: The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018

    2018

  13. [21]

    J. Liu, C. Gao, D. Meng, A. G. Hauptmann, Decidenet: Count- ing varying density crowds through attention guided detection and density estimation, in: The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018

    2018

  14. [22]

    Ranjan, H

    V. Ranjan, H. Le, M. Hoai, Iterative crowd counting, in: The European Conference on Computer Vision (ECCV), 2018

    2018

  15. [23]

    W. Liu, M. Salzmann, P. Fua, Context-aware crowd counting, CoRR abs/1811.10452

    arXiv

  16. [24]

    L.-C. Chen, G. Papandreou, I. Kokkinos, K. Murphy, A. L. Yuille, Deeplab: Semantic image segmentation with deep con- volutional nets, atrous convolution, and fully connected crfs, arXiv preprint arXiv:1606.00915

    arXiv

  17. [25]

    S. S. Kruthiventi, K. Ayush, R. V. Babu, Deepfix: A fully con- volutional neural network for predicting human eye fixations, IEEE Transactions on Image Processing

  18. [26]

    Y. Ji, H. Zhang, Q. J. Wu, Salient object detection via multi- scale attention cnn, Neurocomputing 322 (2018) 130 – 140

    2018

  19. [27]

    A. R. Gepperth, M. G. Ortiz, E. Sattarov, B. Heisele, Dynamic attention priors: a new and efficient concept for improving ob- ject detection, Neurocomputing 197 (2016) 14 – 28

    2016

  20. [28]

    K. G. Y. Y. S. C. Shuangjie Xu, Yu Cheng, P. Zhou, Jointly attentive spatial-temporal pooling networks for video-based per- son re-identification, in: ICCV, 2017

    2017

  21. [29]

    V. Mnih, N. Heess, A. Graves, K. Kavukcuoglu, Recurrent mod- els of visual attention, CoRR abs/1406.6247. URL http://arxiv.org/abs/1406.6247

    arXiv

  22. [30]

    Z. Gan, Y. Cheng, A. E. Kholy, L. Li, J. Liu, J. Gao, Multi- step reasoning via recurrent dual attention for visual dialog, in: Proceedings of the 57th Conference of the Association for Computational Linguistics, ACL 2019, Florence, Italy, July 28- August 2, 2019, Volume 1: Lo...

    2019

  23. [31]

    J. Ba, V. Mnih, K. Kavukcuoglu, Multiple object recognition with visual attention, CoRR abs/1412.7755. URL http://arxiv.org/abs/1412.7755

    arXiv

  24. [32]

    Almahairi, N

    A. Almahairi, N. Ballas, T. Cooijmans, Y. Zheng, H. Larochelle, A. C. Courville, Dynamic capacity networks, CoRR abs/1511.07838

    arXiv

  25. [33]

    Hossain, M

    M. Hossain, M. Hosseinzadeh, O. Chanda, Y. Wang, Crowd counting using scale-aware attention networks, in: 2019 IEEE Winter Conference on Applications of Computer Vision (WACV), 2019, pp. 1280–1288. doi:10.1109/WACV.2019.00141

    2019

  26. [34]

    L. Zhu, Z. Zhao, C. Lu, Y. Lin, Y. Peng, T. Yao, Dual path multi-scale fusion networks with attention for crowd counting, CoRR abs/1902.01115

    1902 arXiv

  27. [35]

    Zhang, H

    C. Zhang, H. Li, X. Wang, X. Yang, Cross-scene crowd counting via deep convolutional neural networks, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2015, pp. 833–841

    2015

  28. [36]

    D. B. Sam, R. V. Babu, Top-down feedback for crowd counting convolutional neural network, CoRR abs/1807.08881

    arXiv

  29. [37]

    A. B. Chan, Z.-S. J. Liang, N. Vasconcelos, Privacy preserving crowd monitoring: Counting people without people models or tracking, in: Computer Vision and Pattern Recognition, 2008. CVPR 2008. IEEE Conference on, IEEE, 2008, pp. 1–7

    2008

  30. [38]

    Kumagai, K

    S. Kumagai, K. Hotta, T. Kurita, Mixture of counting cnns: Adaptive integration of cnns specialized to specific appearance for crowd counting, CoRR abs/1703.09393

    arXiv

  31. [39]

    V. Q. Pham, T. Kozakaya, O. Yamaguchi, R. Okada, Count for- est: Co-voting uncertain number of targets using random forest for crowd density estimation, in: IEEE International Conference on Computer Vision, 2015, pp. 3253–3261

    2015

  32. [40]

    Idrees, I

    H. Idrees, I. Saleemi, C. Seibert, M. Shah, Multi-source multi- scale counting in extremely dense crowd images, in: The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2013

    2013

  33. [41]

    Boominathan, S

    L. Boominathan, S. S. S. Kruthiventi, R. V. Babu, Crowdnet: A deep convolutional network for dense crowd counting, in: Pro- ceedings of the 2016 ACM on Multimedia Conference, MM ’16, ACM, New York, NY, USA, 2016, pp. 640–644

    2016

  34. [42]

    Cheng, Q

    Y. Cheng, Q. Fan, S. Pankanti, A. Choudhary, Temporal se- quence modeling for video event detection, in: Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition, CVPR ’14, 2014, pp. 2235–2242

    2014

  35. [43]

    J. Wang, Y. Cheng, R. Schmidt Feris, Walk and learn: Fa- cial attribute representation learning from egocentric video and contextual data, in: The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016. 10

    2016

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.