Pith. sign in

REVIEW 4 major objections 5 minor 66 references

Crowd Detection Using Very-Fine-Resolution Satellite Imagery

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read CrowdSat-Net reports an F1-score of 66.12% for detecting individual people in 0.3-meter satellite imagery, beating five prior crowd-detection methods.

desk verdict The dataset is the real contribution; the claimed state-of-the-art result rests on a single run with an arbitrary matching threshold, so the method comparison should be treated as provisional. read the letter →

arxiv 2504.19546 v1 pith:2PP33PBU submitted 2025-04-28 cs.CV

Tong Xiao , Qunming Wang , Ping Lu , Tenghai Huang , Xiaohua Tong , Peter M. Atkinson This is my paper · ORCID
classification cs.CV
keywords crowddetectionvery-fine-resolutionsatelliteimagerypoint-basedconvolutionalneuralnetworkSatdatasetFIDTmapsmall-objectremotesensingdeformableconvolution
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 sets out to establish that very-fine-resolution (VFR) satellite imagery, at about 0.3 meters per pixel, can support crowd detection—locating individual people rather than just counting them—a task previously limited to ground and aerial imagery. To do this it introduces CrowdSat, a dataset of over 120,000 manually labeled individuals from 32 Chinese regions collected from the BJ3N and JL4A satellites and the Google Earth platform, and CrowdSat-Net, a point-based convolutional network with two designed components: DCPAN, which sharpens the weak signal of a person (roughly 3×3 pixels), and HFGDU, which restores high-frequency detail lost during upsampling. On CrowdSat, CrowdSat-Net reports the highest F1-score (66.12%) and precision (73.23%) among six compared methods, and it maintains competitive scores on six unseen foreign sites. If the claim is right, large-scale and historical crowd monitoring from space becomes practical.

What carries the argument

The load-bearing objects are the two inserted modules and the training target they serve. DCPAN (Dual-Context Progressive Attention Network) combines base spatial attention with a multi-scale dilated-context branch and a local-contrast branch to produce a per-pixel enhancement weight, counteracting the attenuation of 3-by-3-pixel person signals in early convolution and pooling. HFGDU (High-Frequency Guided Deformable Upsampler) first applies bilinear interpolation, then compensates high frequencies through a learnable Laplacian-like filter and residual refinement, and finally aligns the upsampled and fine features using deformable convolution with an adaptive modulation gate. The FIDT map supplies a smooth regression target whose local maxima mark person centers, and the ablation design isolates each module's contribution: DCPAN alone adds 0.90 F1 and 4.36 precision, HFGDU alone adds 1.10 F1 and 2.09 precision, and together they add 1.70 F1 and 4.69 precision over the baseline.

What would settle it

Re-run all six methods on the CrowdSat validation patches with the matching threshold changed from 1 pixel to 2 and then 3 pixels, leaving training untouched; if CrowdSat-Net stops being the top F1 method, its claimed advantage is an artifact of the chosen tolerance.

Watch

Extended reading notes

Core claim

The central claim is that individual people can be localized point-by-point in sub-meter satellite imagery, provided the network is built around the two ways such small signals die: attenuation during feature extraction and high-frequency blur during upsampling. CrowdSat-Net is a two-stacked Hourglass network trained on FIDT maps with Focal Loss; DCPAN computes an enhancement weight from spatial attention, multi-scale dilated context, and local contrast, then multiplies it into the features; HFGDU replaces bilinear upsampling with a learnable Laplacian high-pass compensation followed by deformable-convolution alignment guided by a modulation gate. The paper reports F1-score of 66.12%, recall of 60.27%, and precision of 73.23% on the CrowdSat validation set, ahead of the second-best method FIDTMCL by 1.71 and 2.42 percentage points in F1 and precision, and ahead of APGCC by 1.78 points in F1 despite APGCC's higher recall. The same model reaches F1-scores of 62.20–75.00% on six foreign landmarks without retraining.

Load-bearing premise

The reported ranking rests on a one-pixel matching tolerance for true positives, and the paper does not test whether a different tolerance changes the result, even though a person spans roughly three by three pixels.

Editorial extensions

If this is right

  • If CrowdSat-Net's result holds, VFR satellite imagery can be used for individual-level crowd localization at national scale, not just crowd counting.
  • The CrowdSat dataset gives ground and aerial crowd-detection methods a new transfer target and a benchmark for small-object localization from space.
  • DCPAN and HFGDU are modular; the paper argues they can be inserted into common coarse-to-fine architectures such as FPN, U-Net, and HRNet to improve small-object detection generally.
  • The method's demonstrated weakness in extremely sparse scenes (precision 35.81%) and extremely dense scenes (recall 39.23%) defines where the next training data and architecture changes are needed.

Reading between the lines

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

  • Because the evaluation uses a matching threshold of one pixel while each person spans about three pixels, comparing all methods under thresholds of two and three pixels would show whether CrowdSat-Net's lead is robust or an artifact of the chosen tolerance; the paper does not report such a sensitivity analysis.
  • CrowdSat's protocol of at most one label per pixel, with only about 50 sub-pixel cases among 120k labels, likely under-represents exactly the extreme-density regime where the method fails; a super-resolved or multi-temporal variant of the dataset would test whether that failure is a data limit or an algorithmic one.
  • The same two failure modes—small-signal attenuation and upsampling blur—appear in other very-fine-resolution satellite tasks such as counting vehicles or wildlife, so DCPAN and HFGDU are plausibly transferable, but transfer needs evaluation on those objects before it can be assumed.
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 presents CrowdSat, a new dataset of very-fine-resolution (VFR) satellite imagery containing over 120,000 manually labeled individuals across China, and CrowdSat-Net, a point-based CNN that adds a Dual-Context Progressive Attention Network (DCPAN) and a High-Frequency Guided Deformable Upsampler (HFGDU) to a stacked Hourglass backbone. The authors report that CrowdSat-Net obtains the highest F1-score (66.12%) and precision (73.23%) among six point-based crowd detection methods on the CrowdSat validation set, with an ablation study attributing the gains to the two new modules. They also demonstrate cross-regional generalization on six scenes outside China and discuss limitations of the dataset and method.

Significance. The dataset is a potentially valuable contribution: it is the first VFR satellite crowd-detection benchmark, covers diverse environments, and the authors release the source code and dataset, which supports reproducibility and follow-up research. The proposed DCPAN and HFGDU modules address plausible failure modes (small-object signal loss and high-frequency detail loss during upsampling) and could transfer to other small-object detection tasks. However, the reported performance advantage is modest (1.71% F1 over the second-best method) and the evaluation protocol has not yet been shown to be robust to the matching threshold or training variability, so the significance of the empirical claims depends on the additional analyses suggested below.

major comments (4)
  1. [Section III.F, Eqs. (14)-(16)] The evaluation uses a matching threshold of tau = 1 pixel with nearest-neighbor matching to define true positives. Because each individual occupies roughly a 3x3-pixel region (Section I), this is an extremely strict localization tolerance, and the paper provides no sensitivity analysis for tau. Table IV shows that the compared methods sit at different precision/recall points (e.g., APGCC has recall 63.05% and precision 65.70%, while CrowdSat-Net has recall 60.27% and precision 73.23%), so a looser threshold, such as tau = 2 or 3 pixels, will likely convert near-miss predictions into true positives at different rates across methods and could change the F1 ranking. The authors should report Precision, Recall and F1 for a range of tau values (e.g., 1, 2, 3 pixels) and state whether the ranking of CrowdSat-Net against FIDTMCL and APGCC is stable.
  2. [Section III.A, Tables I-IV] All reported numbers appear to come from a single training run, with no error bars, multiple seeds, or significance testing. The headline improvement over FIDTMCL is only 1.71 percentage points in F1 (66.12% vs. 64.41%), which could plausibly lie within run-to-run variation for these models. The authors should run each configuration at least three times (with different random seeds) and report means and standard deviations, and ideally a paired significance test, for the main comparison and for the ablations in Tables I-III.
  3. [Section III.D and Fig. 9] The performance analysis for the 'extremely dense' group (800+ individuals) appears to rely on a single validation image, since the dataset maximum count is 808 (Section II.C) and the group is defined by the quantile-based range 800+. The reported Recall drop to 39.23% for this group is therefore a one-sample observation. The authors should either construct more dense validation samples (e.g., by using a different split or collecting additional dense patches) or unambiguously present the result as a case study rather than a general density-dependent finding.
  4. [Section III.A] All compared methods are trained with an identical fixed schedule (150 epochs, Adam with learning rate 0.0003, weight decay 0.001, batch size 8) and no per-method hyperparameter tuning. This protocol may disadvantage methods that need different schedules; for instance, PET's very low F1 of 11.36% in Table IV suggests the model may not have been trained under favorable conditions. The comparison should be made fairer by tuning each baseline on the validation set (or reporting a small grid search), or the authors should provide evidence that the fixed schedule is near-optimal for all compared methods.
minor comments (5)
  1. [Various sections] Section numbering is inconsistent: the Experiments section is labeled 'III' (the same as Methods), Discussion is labeled 'IV' instead of 'V', and Conclusion is labeled 'V' instead of 'VI'.
  2. [Sections IV.B, IV.A, II.A] The module name is spelled inconsistently: 'HFGDU' in most places but 'HFGDO' in Sections IV.B and V; similarly, 'CrowSat' appears in the heading of Section IV.A instead of 'CrowdSat', and the satellite is sometimes 'JL04A' and sometimes 'JL4A'.
  3. [Section III.D] Figure numbering is confusing: the density-performance plot is referred to as Fig. 11 in the text but is captioned Fig. 9, and the visual comparison figure is captioned Fig. 11 but should likely be re-numbered; please renumber all figures consistently.
  4. [Section III.B, Eq. (1)] In Eq. (1) and the surrounding text, the FIDT weight factors and epsilon are defined, but the phrase 'weight' for alpha and beta is not standard; consider using 'weight factors' consistently and defining the role of each constant explicitly.
  5. [Section III.F] The paper would benefit from a short discussion of how the chosen tau compares to typical tolerances in existing crowd localization benchmarks (e.g., FIDTMCL, NWPU-Crowd), since the current citation [48] is from an animal-counting domain.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the reported F1 and Precision are held-out empirical results computed against manually labeled ground-truth points, with no fitted quantity renamed as a prediction.

full rationale

The paper's central claim is an empirical result: CrowdSat-Net is trained on the new CrowdSat dataset and evaluated on held-out validation patches against published point-based baselines, with F1, Precision, and Recall computed from TP/FP/FN after nearest-neighbor matching at tau=1 pixel (Section III.F, Eqs. 14-16). Nothing in the derivation defines the predicted metric in terms of model inputs: the training target (FIDT map from [41], Eq. 1) is a fixed label transform with parameters taken from prior work, not a fit to the validation labels or to the evaluation metric; training loss is Focal Loss on the transformed maps, not the F1 score; and the final points are extracted by LMDS and compared directly to manually labeled ground-truth points. The use of ATPRK pan-sharpening cites prior work by the same group, but it is an externally published preprocessing method and is not the source of the reported superiority. The tau=1 pixel threshold is indeed not sensitivity-analyzed, but that is an evaluation-robustness concern, not a circularity: the threshold is a fixed external rule applied identically to all methods, and changing it would alter all scores without making any method's 'prediction' reduce to its fitted inputs. No step was found in which an output quantity equals an input by construction, a fitted parameter is renamed as a prediction, or the central claim depends on a self-citation chain.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The paper introduces no new physical entities; the novel modules DCPAN and HFGDU are architectural components, not postulated entities.

free parameters (5)
  • LMDS adaptive threshold ratio = 100/255.0 * max FIDT
    Used to filter false positives in local maxima detection; chosen empirically in Section III.E, no sensitivity analysis.
  • LMDS fixed threshold for no detection = 0.10
    If max FIDT < 0.10, no individuals detected; inherited from FIDT [41] but set as constant.
  • Matching distance threshold tau = 1 pixel
    Determines TP/FP/FN in evaluation; set to 1 pixel citing [48], no sensitivity analysis.
  • FIDT alpha = 0.02
    Weight in FIDT map generation; default from [41], not tuned on CrowdSat.
  • FIDT beta = 0.75
    Weight in FIDT map generation; default from [41].
assumptions (3)
  • domain assumption Manual annotations of individual people in 0.3 m satellite imagery are accurate and consistent enough to serve as ground truth.
    The entire evaluation depends on label quality; no inter-annotator agreement study is reported (Section II.B).
  • domain assumption The chosen training hyperparameters (Adam, lr=0.0003, 150 epochs, batch size 8) are appropriate for all compared methods.
    All baselines are trained with the same schedule without per-method tuning (Section IV.A), so the comparison assumes equal suitability.
  • domain assumption The 1-pixel matching threshold is a valid criterion for evaluating detection at the target scale.
    Tau=1 pixel is adopted from [48]; the paper does not test sensitivity to tau, even though objects are only about 3x3 pixels (Section III.F).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Crowd Detection Using Very-Fine-Resolution Satellite Imagery." pith.science (2026). https://pith.science/paper/2PP33PBU

@misc{pith2026250419546,
  author       = {Pith},
  title        = {Pith review of: Crowd Detection Using Very-Fine-Resolution Satellite Imagery},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2PP33PBU}},
  note         = {Machine review of arXiv:2504.19546}
}
read the original abstract

Accurate crowd detection (CD) is critical for public safety and historical pattern analysis, yet existing methods relying on ground and aerial imagery suffer from limited spatio-temporal coverage. The development of very-fine-resolution (VFR) satellite sensor imagery (e.g., ~0.3 m spatial resolution) provides unprecedented opportunities for large-scale crowd activity analysis, but it has never been considered for this task. To address this gap, we proposed CrowdSat-Net, a novel point-based convolutional neural network, which features two innovative components: Dual-Context Progressive Attention Network (DCPAN) to improve feature representation of individuals by aggregating scene context and local individual characteristics, and High-Frequency Guided Deformable Upsampler (HFGDU) that recovers high-frequency information during upsampling through frequency-domain guided deformable convolutions. To validate the effectiveness of CrowdSat-Net, we developed CrowdSat, the first VFR satellite imagery dataset designed specifically for CD tasks, comprising over 120k manually labeled individuals from multi-source satellite platforms (Beijing-3N, Jilin-1 Gaofen-04A and Google Earth) across China. In the experiments, CrowdSat-Net was compared with five state-of-the-art point-based CD methods (originally designed for ground or aerial imagery) using CrowdSat and achieved the largest F1-score of 66.12% and Precision of 73.23%, surpassing the second-best method by 1.71% and 2.42%, respectively. Moreover, extensive ablation experiments validated the importance of the DCPAN and HFGDU modules. Furthermore, cross-regional evaluation further demonstrated the spatial generalizability of CrowdSat-Net. This research advances CD capability by providing both a newly developed network architecture for CD and a pioneering benchmark dataset to facilitate future CD development.

Figures

Figures reproduced from arXiv: 2504.19546 by the authors.

Figure 1
Figure 1. Examples of VFR satellite imagery for CD. (a) Yuesha Island, Inner Mongolia, China, acquired by the Google Earth platform on Feb. 22, 2023 with a spatial resolution of 0.30 m. (b) Shanghai Disney, China, acquired by the JL4A satellite on Feb. 16, 2023 with a spatial resolution of 0.31 m. (c) Harbin Ice and Snow World on Feb. 2, 2024 and (d) the Xi'an Emperor Qinshihuang's Mausoleum Site Museum in China on Oct. 29, 2… view at source ↗
Figure 2
Figure 2. Examples of ground and aerial imagery. (a) image taken from the ShanghaiTech Dataset1 and (b) image obtained from the DLR Aerial Crowd Dataset2 . To overcome the aforementioned limitations, this paper proposed a novel point-based convolutional neural network (CNN) method, CrowdSat-Net, which was specifically designed for large-scale and long-term CD. CrowdSat-Net introduces two key contributions: 1) a Dual-Context P… view at source ↗
Figure 3
Figure 3. Spatial distribution map of collected VFR satellite imagery for some locations where crowds typically gather. To ensure broad spatial coverage and diverse conditions, we selected imagery spanning 32 provincial-level divisions in China (except Guizhou Province and Macao, due to fewer satellite images in these areas). These regions exhibit substantial heterogeneity, encompassing various landscapes and urban settings, … view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Examples of mislabeling. (a) and (b) were captured in the same region on Feb. 16, 2023 from the BJ3N satellite and Aug. 7, 2024 from the Google Earth platform, respectively. The objects in the red and cyan circles correspond to road asphalt and street lamps, respective…
Figure 5
Figure 5. Figure 5: Crowd count distribution in CrowdSat. Extensive National Coverage. CrowdSat spans China (except Guizhou Province and Macao) and includes samples from 32 Chinese provincial-level divisions. This vast coverage provides a wide range of crowd types, promoting the generaliz…
Figure 6
Figure 6. Figure 6: Overview of the proposed CrowdSat-Net. First, the labeled image is transformed into the FIDT map. During each training iteration, CrowdSat-Net enhances the basic features in the image preprocessing using the DCPAN module. Then, these enhanced features pass through the …
Figure 7
Figure 7. Figure 7: Flowchart of the DCPAN module, which consists of three main components: SA Encoding, MSFE and LCE branches. The SA Encoding component extracts the base SA feature from the original features X. The MSFE branch captures contextual information from this base feature, whil…
Figure 8
Figure 8. Figure 8: The structure of the HFGDU module. It contains three main components: Initial Upsampling, HFDC and DAF. Initial Upsampling uses the bilinear interpolation to match the spatial resolution of coarse and fine features. Then, HFDC recovers the high-frequency details of the…
Figure 9
Figure 9. Figure 9: Performance of CrowdSat-Net on CrowdSat across different crowd densities. The lines show the Precision, Recall and F1-score evaluated on subsets of images grouped by crowd densities, while the grey bars in the background indicate the number of images in each group. E. …
Figure 10
Figure 10. Figure 10: Examples of the localization performance of CrowdSat-Net in extreme scenarios: (a) Extremely sparse. (b) Extremely dense [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]
Figure 11
Figure 11. Figure 11: Visual comparisons between different CD methods for the CrowdSat dataset. (a) Original image. (b) Reference. (c) SCALNet. (d) P2PNet. (e) PET. (f) FIDTMCL. (g) APGCC. (h) CrowdSat-Net. (I) Traffic junctions. (II) Snowfields. (III) Dense urban regions. (IV) Desert regi…
Figure 12
Figure 12. Figure 12: The visual localization performance of CrowdSat-Net in unseen foreign regions. (a) Phra Nakhon. (b) India Gate. (c) National Mall. (d) Djemaa el Fna. (e) Metropolitan Cathedral. (f) Red Square. (I) Original image. (II) Prediction. (III) Evaluation [PITH_FULL_IMAGE:fi…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

66 extracted references · 38 canonical work pages

  1. [1]

    Reducing traffic congestion in makkah during Hajj through the use of AI technology,

    F. Gazzawe and M. Albahar, “Reducing traffic congestion in makkah during Hajj through the use of AI technology,” Heliyon, vol. 10, no. 1, 2024

    work page 2024

  2. [2]

    Ranking and analysis the strategies of cro wd management to reduce the risks of crushes and stampedes in crowded environments and ensure the safety of passengers,

    W. T. Al -Nami, “Ranking and analysis the strategies of cro wd management to reduce the risks of crushes and stampedes in crowded environments and ensure the safety of passengers,” Neutrosophic Systems with Applications, vol. 8, pp. 61–78, 2023

    work page 2023

  3. [3]

    Felic iani, K

    C. Felic iani, K. Shimura, and K. Nishinari, Introduction to Crowd Management: Managing Crowds in the Digital Era: Theory and Practice . Cham: Springer International Publishing, 2021. doi: 10.1007/978 -3-030- 90012-0

    doi:10.1007/978 2021

  4. [4]

    A Literature Revie w on Impact of COVID - 19 Pandemic on Teaching and Learning,

    S. Pokhrel and R. Chhet ri, “A Literature Revie w on Impact of COVID - 19 Pandemic on Teaching and Learning, ” Higher Education for the Future, vol. 8, no. 1, pp. 133 –141, Jan. 2021, doi: 10.1177/2347631120983481

    doi:10.1177/2347631120983481 2021

  5. [5]

    Problematising density: COVID -19, the crowd, and urban life,

    A. Joiner, C. McFarlane, L. Rella, and M. Uriarte -Ruiz, “Problematising density: COVID -19, the crowd, and urban life,” Social & Cultural Geography, vol. 25, no. 2, pp. 181–198, 2024

    work page 2024

  6. [6]

    Locate, Size and Count: Accurately Resolving People in Dense Crowds via Detection

    D. B. Sam, S. V. Peri, M. N. Sundararaman, A. Kamath, and R. V. Babu, “Locate, Size and Count: Accurately Resolving People in Dense Cro wds via Detection,” Feb. 15, 2020, arXiv:1906.07538

    work page Pith review arXiv 2020

  7. [7]

    A Generalized Loss Function for Crowd Counting and Localization,

    J. Wan, Z. Liu, and A. B. Chan, “A Generalized Loss Function for Crowd Counting and Localization,” in 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , Nashville, TN, USA: IEEE, Jun. 2021, pp. 1974–1983. doi: 10.1109/CVPR46437.2021.00201

    arXiv 2021

  8. [8]

    STEERER: Resolving Scale Variations for Counting and Localization via Selective Inheritance Learning

    T. Han, L. Bai, L. Liu, and W. Ouyang, “STEERER: Resolving Scale Variations for Counting and Localization via Selective Inheritance Learning,” Aug. 21, 2023, arXiv:2308.10468

    work page Pith review arXiv 2023

Show all 66 references
  1. [9]

    Review of analyses on crowd-gathering risk a nd its evaluation methods,

    W. W eng, J. Wang, L. Shen, and Y. Song, “Review of analyses on crowd-gathering risk a nd its evaluation methods,” Journal of safety science and resilience, vol. 4, no. 1, pp. 93–107, 2023

    2023

  2. [10]

    A survey on deep learning -based real -time crowd anomaly detection for secure distrib uted video surveillance,

    K. Rezaee, S. M. Rezakhani, M. R. Khosravi, and M. K. Moghimi, “A survey on deep learning -based real -time crowd anomaly detection for secure distrib uted video surveillance,” Personal and Ubiquitous Computing, vol. 28, no. 1, pp. 135–151, 2024

    2024

  3. [11]

    Correlating urban spatial form and crowd s patiotemporal behavior: A case study of Lhasa, China,

    Z. Luo, L. Marchi, F. Chen, Y. Zhang, and J. Gaspari, “Correlating urban spatial form and crowd s patiotemporal behavior: A case study of Lhasa, China,” Cities, vol. 160, p. 105812, 2025

    2025

  4. [12]

    CountFormer: Multi-view Crowd Counting Transformer,

    H. Mo et al., “CountFormer: Multi-view Crowd Counting Transformer, ” in European Conference on Computer Vision, Springer, 2024, pp. 20–40

    2024

  5. [13]

    Learning to count objects in images,

    L. Victor and Z. A ndrew, “Learning to count objects in images,” Advances in neural information processing systems, vol. 23, 2010

    2010

  6. [14]

    Cross -view cross-scene multi-view crowd counting,

    Q. Zhang, W. Lin, and A. B. Chan, “Cross -view cross-scene multi-view crowd counting,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 557–567

    2021

  7. [15]

    MRCNet: Crowd Counting and Density Map Estimation in Aerial and Ground Imagery,

    R. Bahmanyar, E. Vig, and P. Reinartz, “MRCNet: Crowd Counting and Density Map Estimation in Aerial and Ground Imagery,” Sep. 27, 2019, arXiv:1909.12743

    2019 arXiv

  8. [16]

    Detection of High -Density Crowds in Aerial Images Using Te xture Classification,

    O. Meynberg, S. Cui, and P. Reinartz, “ Detection of High -Density Crowds in Aerial Images Using Te xture Classification,” Remote Sensing, vol. 8, no. 6, p. 470, Jun. 2016, doi: 10.3390/rs8060470

    2016 doi

  9. [17]

    Abnormal crowd density estimation in aerial images,

    H. Mliki, A. Olfa, and H. Mohamed, “Abnormal crowd density estimation in aerial images,” Journal of Electronic Imaging , vol. 28, no. 1, pp. 013047–013047, 2019

    2019

  10. [18]

    Regulations of Hebei Province on Aerial Surveyi ng and Photogrammetry,

    The People's Government of Hebei Province, “Regulations of Hebei Province on Aerial Surveyi ng and Photogrammetry,” The People's Government of Hebei Province , https://www.hebei.gov.cn/columns/67c6d89b-b5b0-4492-98d6- 997dac63a29b/202309/12/0808f9e0-84c9-4580-82e9-bfe3b734c875...

    2011

  11. [19]

    Counting Dense Objects in Remote Sensing Images,

    G. Gao, Q. Liu, and Y. W ang, “Counting Dense Objects in Remote Sensing Images,” in ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (I CASSP), 2020, pp. 4137 –

    2020

  12. [20]

    Counting From Sky: A Large -Scale Data Set for Remote Sensing Object Counting and a Benchmark Method,

    G. Gao, Q. Liu, and Y. Wang, “ Counting From Sky: A Large -Scale Data Set for Remote Sensing Object Counting and a Benchmark Method,” IEEE Trans. Geosci. Remote Sensing, vol. 59, no. 5, pp. 3642–3655, May 2021, doi: 10.1109/TGRS.2020.3020555

    2021

  13. [21]

    A Lightweight Multiscale Feature Fusion Network for Remote Sensing Object Counting,

    J. Yi, Z. Shen, F. Chen, Y. Zhao, S. Xia o, and W. Zhou, “A Lightweight Multiscale Feature Fusion Network for Remote Sensing Object Counting,” IEEE Transactions on Geoscience and Remote Sensing , vol. 61, pp. 1–13, 2023, doi: 10.1109/TGRS.2023.3238185

    2023

  14. [22]

    Quality assessment of fusing Sentinel-2 and WorldView-4 imagery on Sentinel-2 spectral band values: a case study of Zagreb, Croatia,

    L. Rumora, M. Gašparović, M. Miler, and D. Medak, “Quality assessment of fusing Sentinel-2 and WorldView-4 imagery on Sentinel-2 spectral band values: a case study of Zagreb, Croatia,” International Journal of Image and Data Fusion , vol. 11, no. 1, pp . 77–96, Jan. 2020, doi:...

    2020

  15. [23]

    Crowd counting in the frequency domain,

    W. Shu, J. Wan, K. C. Tan, S. Kwong, and A. B. Chan, “Crowd counting in the frequency domain,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 19618–19627

    2022

  16. [24]

    Revisiting Crowd Counting: State-of-the-art, Trends, and Future Perspectives,

    M. A. Khan, H. Menouar, and R. Hamila, “Revisiting Crowd Counting: State-of-the-art, Trends, and Future Perspectives,” Sep. 14, 2022, arXiv:2209.07271

    2022 arXiv

  17. [25]

    A Perspective - Embedded Scale -Selection Network for Crowd Counting in Public Transportation,

    J. Yi, Y. Pang, W. Zhou, M. Zhao, and F. Zheng, “A Perspective - Embedded Scale -Selection Network for Crowd Counting in Public Transportation,” IEEE Transactions on Intelligent Transportation Systems, vol. 25, no. 5, pp. 3420 –3432, 2024, doi: 10.1109/TITS.2023.3328000

    2024

  18. [26]

    Where Are the Blobs: Counting by Localization with Point Supervision,

    I. H. Laradji, N. Ros tamzadeh, P. O. Pinheiro, D. Vazquez, and M. Schmidt, “Where Are the Blobs: Counting by Localization with Point Supervision,” in Computer Vision – ECCV 2018, V. Ferrari, M. Hebert, C. Sminchisescu, and Y. Weiss, Eds., Cham: Springer International Publishi...

    2018

  19. [27]

    Re thinking Counting and Localization in Crowds: A Purely Point -Based Framework,

    Q. Song et al. , “Re thinking Counting and Localization in Crowds: A Purely Point -Based Framework,” in 2021 IEEE/CVF International Conference on Computer Vision (ICCV) , Montreal, QC, Canada: IEEE, Oct. 2021, pp. 3345–3354. doi: 10.1109/ICCV48922.2021.00335

    2021

  20. [28]

    Density map regression guided detection network for rgb -d crowd counting and localization,

    D. Lian, J. Li , J. Zheng, W. Luo, and S. Gao, “Density map regression guided detection network for rgb -d crowd counting and localization,” presented at the Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019, pp. 1821--1830

    2019

  21. [29]

    Recurrent Attentive Zooming for Joint Crowd Counting and Precise Localization,

    C. L iu, X. Weng, and Y. Mu, “Recurrent Attentive Zooming for Joint Crowd Counting and Precise Localization,” in Proceedings of the 17 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Jun. 2019

    2019

  22. [30]

    Gaussian Process Density Counting from Weak Supervision,

    M. von Borstel, M. Kandemir, P. Schmidt, M . K. Rao, K. Rajamani, and F. A. Hamprecht, “Gaussian Process Density Counting from Weak Supervision,” in Computer Vision – ECCV 2016, B. Leibe, J. Matas, N. Sebe, and M. Welling, Eds., Cham: Springer Internati onal Publishing, 2016, ...

    2016

  23. [31]

    Faster r -cnn: Towards real-time object detection with region proposal networks,

    S. Ren, K. He, R. Girshick, and J. Sun, “Faster r -cnn: Towards real-time object detection with region proposal networks,” Advances in neural information processing systems, vol. 28, 2015

    2015

  24. [32]

    Objects as Points,

    X. Zhou, D. Wang, and P. Krähenbühl, “Objects as Points,” Apr. 25, 2019, arXiv:1904.07850

    2019 arXiv

  25. [33]

    Tracking Objects as Points,

    X. Zhou, V. Koltun, and P. Krähenbühl, “Tracking Objects as Points,” Aug. 21, 2020, arXiv:2004.01177

    2020 arXiv

  26. [34]

    A survey and performance evaluation of deep learning methods for small objec t detection,

    Y. Liu, P. Sun, N. Wergeles, and Y. Shang, “A survey and performance evaluation of deep learning methods for small objec t detection,” Expert Systems with Applications , vol. 172, p. 114602, 2021, doi: https://doi.org/10.1016/j.eswa.2021.114602

    2021

  27. [35]

    Deep learning -based detection from the perspective of small o r tiny objects: A survey,

    K. Tong and Y. Wu, “Deep learning -based detection from the perspective of small o r tiny objects: A survey,” Image and Vision Computing, vol. 123, p. 104471, 2022, doi: https://doi.org/10.1016/j.imavis.2022.104471

    2022

  28. [36]

    A review of small object detection based on deep learning,

    W. Wei, Y. Cheng, J. He, and X. Zhu, “A review of small object detection based on deep learning,” Neural Computing and Applications , vol. 36, no. 12, pp. 6283 –6303, Apr. 2024, doi: 10.1007/s00521 -024- 09422-6

    2024 doi

  29. [37]

    Sfnet: Faster and accurate semantic segmentation via semantic flow,

    X. Li et al. , “Sfnet: Faster and accurate semantic segmentation via semantic flow,” International Journal of Computer Vision , vol. 132, no. 2, pp. 466–489, 2024

    2024

  30. [38]

    Deformable Convolutiona l Networks,

    J. Dai et al. , “Deformable Convolutiona l Networks,” in Proceedings of the IEEE International Conference on Computer Vision (ICCV), 2017

    2017

  31. [39]

    Downscaling MODIS images with area -to-point regression kriging,

    Q. Wang, W. Shi, P. M. Atkinson, and Y. Zhao, “Downscaling MODIS images with area -to-point regression kriging,” Remote Sensing of Environment, vol. 166, pp. 191 –204, Sep. 2015, doi: 10.1016/j.rse.2015.06.003

    2015 doi

  32. [40]

    Fusion of Sentinel -2 images,

    Q. Wang, W. Shi, Z. Li, and P . M. Atkinson, “Fusion of Sentinel -2 images,” Remote Sensing of Environment , vol. 187, pp. 241 –252, Dec. 2016, doi: 10.1016/j.rse.2016.10.030

    2016 doi

  33. [41]

    Focal Inverse Distance Transform Maps for Crowd Localization,

    D. Liang, W. Xu, Y. Zhu, and Y. Zhou, “Focal Inverse Distance Transform Maps for Crowd Localization,” IEEE Trans. Multimedia, vol. 25, pp. 6040–6052, 2023, doi: 10.1109/TMM.2022.3203870

    2023

  34. [42]

    Stacked Hourglass Networks for Human Pose Estimation,

    A. Newell, K. Yang, and J. Deng, “Stacked Hourglass Networks for Human Pose Estimation,” in Computer Vision – ECCV 2016, B. Leibe, J. Matas, N. Sebe, and M. Welling, Eds., Cham: Springer International Publishing, 2016, pp. 483–499

    2016

  35. [43]

    Focal loss for dense object detection,

    T.-Y. Lin, P. Goyal, R. Girshick, K. He, and P. Dollár, “Focal loss for dense object detection,” in Proceedings of the IEEE international conference on computer vision, 2017, pp. 2980–2988

    2017

  36. [44]

    AutoScale: Learning to Scale for Crowd Counting,

    C. Xu et al. , “AutoScale: Learning to Scale for Crowd Counting,” International Journal of Computer Vision , vol. 130, no. 2, pp. 405 –434, Feb. 2022, doi: 10.1007/s11263-021-01542-z

    2022 doi

  37. [45]

    Localization in the Crowd with Topological Constraints,

    S. Abousamra, M. Hoai, D. Samaras, and C. Chen, “Localization in the Crowd with Topological Constraints,” in AAAI Conference on Artificial Intelligence (AAAI), 2021

    2021

  38. [46]

    Learning Independent Instance Maps for Crowd Localization,

    J. Gao, T. Han, Y. Yuan, and Q. Wang, “Learning Independent Instance Maps for Crowd Localization,” arXiv:2012.04164, 2020

    2012 arXiv

  39. [47]

    Dual Attention Network for Scene Segmentation,

    J. Fu et al., “Dual Attention Network for Scene Segmentation,” in 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019, pp. 3141–3149. doi: 10.1109/CVPR.2019.00326

    2019

  40. [48]

    From crowd to herd counting: How to precisely detect and count African mammals using aerial imagery and deep learning?,

    A. Delplanque, S. Foucher, J. Thé au, E. Bussiè re, C. Vermeulen, and P. Lejeune, “From crowd to herd counting: How to precisely detect and count African mammals using aerial imagery and deep learning?,” ISPRS Journal of Photogrammetry and Remote Sensing, vol. 197, pp. 167 –18...

    2023 doi

  41. [49]

    K -Nearest Neighbors,

    O. Kramer, “K -Nearest Neighbors,” in Dimensionality Reduction with Unsupervised Nearest Neighbors , Berlin, Heidelberg: Springer Berlin Heidelberg, 2013, pp. 13–23. doi: 10.1007/978-3-642-38652-7_2

    2013 doi

  42. [50]

    NWPU -Crowd: A Large -Scale Benchmark for Crowd Counting and Localization,

    Q. Wa ng, J. Gao, W. Lin, and X. Li, “NWPU -Crowd: A Large -Scale Benchmark for Crowd Counting and Localization,” IEEE Trans. Pattern Anal. Mach. Intell. , vol. 43, no. 6, pp. 2141 –2149, Jun. 2021, doi: 10.1109/TPAMI.2020.3013269

    2021

  43. [51]

    Cutmix: Regularization strategy to train strong classifiers with localizable features,

    S. Yun, D. Han, S. J. Oh, S. Chun, J. Choe, and Y. Yoo, “Cutmix: Regularization strategy to train strong classifiers with localizable features,” in Proceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 6023–6032

    2019

  44. [52]

    Adam: A metho d for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A metho d for stochastic optimization,” arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  45. [53]

    Residual Attention Network for Image Classification,

    F. Wang et al., “Residual Attention Network for Image Classification,” in 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017, pp. 6450–6458. doi: 10.1109/CVPR.2017.683

    2017 doi

  46. [54]

    ECA -Net: Efficient Channel Attention for Deep Convolutional Neural Networks,

    Q. Wang, B. Wu, P. Zhu, P. Li, W. Zuo, and Q. Hu, “ECA -Net: Efficient Channel Attention for Deep Convolutional Neural Networks,” in 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA: IEEE, Jun. 2020, pp. 11531 –11539. doi: 10.1109/CV...

    2020

  47. [55]

    CARAFE: Content-Aware ReAssembly of FEatures,

    J. Wang, K. Chen, R. Xu, Z. Liu, C. C. Loy, and D. Lin, “CARAFE: Content-Aware ReAssembly of FEatures,” in 2019 IEEE/CVF International Conference on Comput er Vision (ICCV) , 2019, pp. 3007 –

    2019

  48. [56]

    Fully Convolutional Networks for Semantic Segmentation,

    E. Shelhamer, J. Long, and T. Darrell, “Fully Convolutional Networks for Semantic Segmentation,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 39, no. 4, pp. 6 40–651, 2017, doi: 10.1109/TPAMI.2016.2572683

    2017

  49. [57]

    Learning to Upsample by Learning to Sample,

    W. Liu, H. Lu, H. Fu, and Z. Cao, “Learning to Upsample by Learning to Sample,” in Proc. IEEE/CVF International Conference on Computer Vision (ICCV), 2023

    2023

  50. [58]

    Dense Poi nt Prediction: A Simple Baseline for C rowd Counting and Localization,

    Y. Wang, X. Hou, and L. -P. Chau, “Dense Poi nt Prediction: A Simple Baseline for C rowd Counting and Localization,” in 2021 IEEE International Conference on Multimedia & Expo Workshops (ICMEW) , 2021, pp. 1–6. doi: 10.1109/ICMEW53276.2021.9455954

    2021

  51. [59]

    Point -Query Q uadtree for Crowd Counting, Localization, and More,

    C. Liu, H. Lu, Z. Cao, and T. Liu, “Point -Query Q uadtree for Crowd Counting, Localization, and More,” Aug. 26, 2023, arXiv:2308.13814

    2023 arXiv

  52. [60]

    Improving Point -Based Crowd Counting and Localization Based on Auxiliary Point Guidance,

    I.-H. Chen, W. -T. Chen, Y. -W. Liu, M. -H. Yang, and S. -Y. Kuo, “Improving Point -Based Crowd Counting and Localization Based on Auxiliary Point Guidance,” in European Conference on Computer Vision, 2024, pp. 428–444

    2024

  53. [61]

    Feature pyramid networks for object detection,

    T.-Y. Lin, P. Dollá r, R. Girshick, K. He, B. Hariharan, and S. Belongie, “Feature pyramid networks for object detection,” in Proceedings of the IEEE conference on computer vision and pattern recogni tion, 2017, pp. 2117–2125

    2017

  54. [62]

    U -Net: Convolutional Networks for Biomedical Image Segmentation,

    O. Ron neberger, P. Fischer, and T. Brox, “U -Net: Convolutional Networks for Biomedical Image Segmentation,” CoRR, vol. abs/1505.04597, 2015

    2015 arXiv

  55. [63]

    Deep high -resolution representation lea rning for human pose estimation,

    K. Sun, B. Xiao, D. Liu, and J. Wang, “Deep high -resolution representation lea rning for human pose estimation,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2019, pp. 5693–5703

    2019

  56. [64]

    Responsible Artificial Intelligence for Earth Observation: Achievable and realistic paths to serve the collective good,

    P. Ghamisi et al., “Responsible Artificial Intelligence for Earth Observation: Achievable and realistic paths to serve the collective good,” IEEE Geoscience and Remote Sensing Magazine, pp. 2–26, 2025, doi: 10.1109/MGRS.2025.3529726

    2025

  57. [3016]

    doi: 10.1109/ICCV.2019.00310

    2019

  58. [4141]

    doi: 10.1109/ICASSP40776.2020.9053690

    2020

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.