Pith. sign in

REVIEW 5 major objections 5 minor 23 references

LRSAA: Large-scale Remote Sensing Image Target Recognition and Automatic Annotation

T0 review · 5 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read LRSAA, an ensemble of YOLOv11 and MobileNetV3-SSD with Poisson-disk tiling and EIOU merging, outperforms five standard detectors in automatic annotation of large urban remote sensing images.

desk verdict Load-bearing coordinate remapping and evaluation contradictions make this engineering pipeline hard to trust as written, though the crop-based detection direction is sound. read the letter →

arxiv 2411.15808 v4 pith:MALLLV5N submitted 2024-11-24 cs.CV

Wuzheng Dong , Yujuan Zhu , Sheng Zhang This is my paper · ORCID
classification cs.CV
keywords remotesensingtargetrecognitionautomaticannotationensemblelearningYOLOv11MobileNetV3-SSDPoissondisksamplingEIOUlosslarge-scaleimagery
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 establish that a practical pipeline for recognizing and automatically labeling objects in very large remote sensing images can beat standard detectors by combining two existing models rather than designing a new architecture. The LRSAA pipeline cuts large aerial images into tiles with Poisson disk sampling, runs YOLOv11 and MobileNetV3-SSD on each tile, merges the detections using an EIOU-based non-maximum suppression, and remaps the surviving boxes to the original image. On 0.6-meter-resolution urban imagery of three cities, the authors report accuracy and F1 scores near 0.79 with 640-pixel tiles, ahead of RetinaNet, SSD VGG, Faster R-CNN, and the two base detectors; they also report that mixing in a proportion of the model's own auto-annotated results as synthetic training data improves recognition. If correct, the contribution is a reusable recipe for large-scale automatic annotation rather than a new detector.

What carries the argument

The load-bearing mechanism is a four-step pipeline: (A) Poisson disk sampling partitions a large image into uniformly distributed, minimally overlapping tiles; (B) YOLOv11 and MobileNetV3-SSD are trained on small tiles and detect objects in each tile; (C) the detected boxes are remapped to original-image coordinates by a matrix offset; and (D) the auto-annotated results are mixed back into the training set as synthetic data and the models are retrained. The identity that carries the merging step is EIOU, an intersection-over-union measure augmented with center-point distance and aspect-ratio terms, used here in non-maximum suppression to discard duplicate boxes more selectively. The ensemble step is result-level: the outputs of both detectors are combined before suppression, so the method's gain comes from fusion, tiling, and remapping rather than from a new detector.

What would settle it

Inspect the released code and run it on one 6400-by-6400 annotated urban tile, printing the remapped boxes before any IoU is computed; if a substantial fraction of mapped boxes have zero width or height, the accuracy, F1, and mAP values in the comparison tables cannot be reproduced and the central outperformance claim collapses.

Watch

Extended reading notes

Core claim

The central claim is that the LRSAA ensemble—YOLOv11 plus MobileNetV3-SSD, with Poisson disk sampling for tile extraction and EIOU-based non-maximum suppression for merging—produces more accurate object detection and automatic annotation on large-scale urban remote sensing imagery than any individual detector or standard single-scale baseline. The paper supports this with comparative tables in which LRSAA at 640-cut achieves the highest average accuracy (0.791) and F1 (0.777) across Tianjin, Shanghai, and Xiamen, and at 320-cut achieves the highest average mAP (0.72). The authors further claim that retraining on a mix of real data and a proportion of automatically annotated synthetic data improves accuracy, F1, and mAP, and that the pipeline reaches this accuracy while keeping computational demand low through tiling.

Load-bearing premise

The entire comparison depends on the coordinate remapping from tile to full image being implemented correctly; the matrix printed in the paper would set every mapped box's maximum coordinate equal to its minimum coordinate plus the tile offset, making every mapped box degenerate, and no code-level evidence is provided that the real implementation avoids this.

Editorial extensions

If this is right

  • If LRSAA works as claimed, automatic annotation of very large remote sensing images can be built from two standard detectors plus tiling, without designing a purpose-built network.
  • The reported gains at 320- and 640-pixel tiles imply that smaller tiles, not larger context, drive the accuracy improvement, which suits memory-constrained deployment.
  • Because the ensemble is modular, the authors expect that further detectors can be added without changing the rest of the pipeline, making the framework extensible.
  • The reported synthetic-data results imply that an auto-annotation system can partially bootstrap its own training data, reducing manual labeling effort.

Reading between the lines

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

  • If the coordinate remapping in the released code matches the printed matrix, the IoU-based numbers in the comparison tables would be invalid, because the printed matrix degenerates every mapped box; a code-level check would be the first thing to try.
  • The 'synthetic data' here is not synthetic imagery but model-generated annotations on real imagery, so the reported improvement is really a self-training or pseudo-labeling effect that could be tested with confidence thresholds.
  • The design isolates three variables—ensemble, tiling, and EIOU merging—so a simple ablation with a single detector plus tiling and standard NMS instead of EIOU would reveal which component actually carries the gain.
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

5 major / 5 minor

Summary. The paper proposes LRSAA, an ensemble of YOLOv11 and MobileNetV3-SSD for object recognition and automatic annotation in large-scale remote sensing images. The pipeline splits large images into crops using Poisson disk sampling, detects objects with the two models, fuses results via EIoU-based NMS, maps detections back to the original image, and then retrains on a mixture of original and synthetic data. Experiments compare LRSAA against RetinaNet, SSD VGG, Faster R-CNN, YOLOv11, and MobileNetV3-SSD on 6400x6400 urban images from Tianjin, Shanghai, and Xiamen, reporting accuracy, F1, and mAP. The paper claims that LRSAA outperforms all competing models, especially with 640-cut and 320-cut images, and that adding synthetic data improves performance.

Significance. If the results held, the paper would offer a practical recipe for large-image remote sensing detection with reduced memory requirements, and the public code repository would aid reproducibility. The use of two complementary detectors, Poisson disk sampling for crop selection, and synthetic-data retraining are all reasonable engineering ideas, and the three-city benchmark with manual annotations is a useful evaluation setup. However, the current manuscript does not sufficiently substantiate these claims: the central coordinate-remapping equation is degenerate, the headline comparison is internally inconsistent, and the synthetic-data experiment has a leakage risk. The strengths are the availability of source code and the attempt to evaluate on real 0.6 m resolution urban imagery, but these do not compensate for the missing technical details and unverified correctness of the mapping step.

major comments (5)
  1. [Section III, coordinate remapping] The matrix equation for mapping detections from crop coordinates back to the original image is degenerate: the third row is identical to the first and the fourth to the second, so a literal reading sets x_max = x'_min + w*nx and y_max = y'_min + h*ny, forcing every mapped box to have zero width and height. If the implementation follows the printed equation, the IoU-based metrics in Table I would be meaningless; if it does not, the paper omits the exact transformation on which the central claim depends. The authors must provide the corrected matrix or the exact code used for remapping.
  2. [Section IV-D, Table I] The claim that 'The proposed LRSAA model outperforms all other models' is contradicted by Table I: on Xiamen with 640-cut images, YOLOv11 achieves Accuracy 0.794, while LRSAA achieves 0.781; on Shanghai, MobileNetV3-SSD 640-cut achieves mAP 0.62 versus LRSAA 0.70, but on Xiamen LRSAA 320-cut mAP 0.72 is tied with YOLOv11 0.69 and MobileNetV3-SSD 0.63, while accuracy is lower than YOLOv11. The comparison should be stated per metric and per city, and the headline claim must be qualified to the settings where it actually holds.
  3. [Section IV-E, Table II] The conclusion that 'there is a significant trend toward performance improvement as the proportion of synthetic data increases' is not supported by the table: the 80% synthetic row has Accuracy 0.716, lower than the 60% row's 0.718, and mAP 0.66 versus 0.67; the 40% row has F1 0.698, below the 0% row's 0.706. No error bars or significance tests are reported. Furthermore, the synthetic data are generated from automatically annotated Tianjin images and the model is then evaluated on manually annotated Tianjin images, which is a within-city pseudo-label leakage that artificially inflates the apparent benefit. A cross-city evaluation or a clear separation between training and evaluation cities is needed.
  4. [Section III, EIoU and fusion details] The EIoU equation is printed as 'EIoU = 1 - IoU + ρ2 + v' with ρ2 and v never defined, and the relationship between EIoU as a bounding-box regression loss and its use in NMS is not explained. The paper also omits the ensemble fusion weights, the NMS threshold, and the procedure for combining YOLOv11 and MobileNetV3-SSD outputs. These details are essential for reproducing the claimed improvements.
  5. [Section IV-C/D, experimental methodology] Table I reports accuracy, F1, and mAP without error bars, number of runs, or statistical significance tests, and the training details (epochs, learning rate, batch size, data augmentation, confidence threshold other than 0.25, IoU threshold definition in 'exceeds 50') are absent. The use of classification accuracy with TN for object detection is also not justified, because true negatives are not well-defined for detection tasks. These omissions make it impossible to assess whether the differences in Table I are meaningful.
minor comments (5)
  1. [Section III, typos] The phrase 'YOLOv11 and MobileNetV3-SSD trainging' contains a typo; 'training' is intended.
  2. [Table II] The header 'Acurracy F1-score mAP' misspells 'Accuracy'.
  3. [Section II-B and References] References [15] and [16] both cite the same Cook 1986 paper; one of the two should be removed or replaced with the Bridson reference that the text actually describes.
  4. [Section IV-B] The text says 'A result is considered accurate if its IoU value exceeds 50'; this should be 'exceeds 0.5' or 'exceeds 50%'.
  5. [Table III, Appendix] Table III appears to be a set of detection visualizations, but no explanation is given for the color coding or the selection of examples, which limits its usefulness as evidence.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the performance claims are empirical, evaluated against manual annotations, and no load-bearing equation reduces to its own input.

full rationale

The paper's derivation chain is empirical rather than definitional. LRSAA is trained on XView and evaluated against manually annotated urban images from Tianjin, Shanghai, and Xiamen, with comparisons to standard detectors such as RetinaNet, SSD VGG, Faster R-CNN, YOLOv11, and MobileNetV3-SSD. No equation in the paper is shown to reduce to its inputs by construction, and there are no self-citations or imported uniqueness theorems that force the method's choice. The coordinate-remapping matrix in Section III is printed with rows [1 0 0 0] and [0 1 0 0] repeated, which would make x_max equal x_min and y_max equal y_min, forcing IoU to zero; this is a serious reproducibility and correctness defect in the claimed evaluation, but it is an inconsistency rather than a circular equivalence, because the reported nonzero metrics imply the implementation does not follow the printed equation. Similarly, the use of model-generated Tianjin pseudo-labels as additional training data is a potential data-leakage or self-training concern, not a definitional circularity, since the final evaluation is against manual annotations. Under the stated criteria, no circular step can be exhibited, so the circularity score is 0.

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

The method has no new physical or algorithmic entities; it combines published detectors and sampling methods. The central claim rests on several unstated implementation choices (Poisson radius, fusion rule, coordinate remapping) and on the reliability of pseudo-labels, none of which are independently verified.

free parameters (5)
  • Confidence threshold = 0.25
    Used in Table I to define accepted detections; chosen without sensitivity analysis.
  • IoU threshold = 0.5
    A detection is counted accurate if IoU with ground truth exceeds 0.5; standard but still a manually chosen decision boundary.
  • Crop size = 640x640 (best of 320/640/1280)
    The paper reports all three and highlights 640-cut as the best average; selecting the best-performing crop size post hoc inflates reported performance.
  • Poisson disk minimum distance r = not reported
    Controls the segmentation density and is never specified, making the exact sampling procedure unreproducible.
  • Ensemble fusion weights/NMS strategy = not specified
    The combination of YOLOv11 and MobileNetV3-SSD outputs is described only as 'ensemble learning' with enhanced NMS; the exact merging rule is not given.
assumptions (5)
  • standard math Bridson's algorithm produces a valid Poisson disk sample with minimum distance r.
    Invoked in Section III as the foundation for the segmentation step.
  • domain assumption XView labels are accurate and representative for training remote sensing detectors.
    Initial training in Section IV-A uses XView without any discussion of label noise or domain shift.
  • domain assumption The manually annotated urban images are correct ground truth.
    Evaluation in Section IV-C assumes the 20 manually annotated regions per city are error-free; no annotation protocol or quality check is given.
  • domain assumption Model-generated pseudo-labels on Tianjin images are reliable enough to use as training data.
    Section IV-E retrains on 'synthetic data' generated from automatically annotated Tianjin images; the reliability of these labels is not verified.
  • domain assumption The coordinate remapping from cropped patches to the original image is implemented correctly.
    The matrix printed in Section III appears to map xmax using the same row as xmin; the paper implicitly assumes the actual code avoids this typo.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LRSAA: Large-scale Remote Sensing Image Target Recognition and Automatic Annotation." pith.science (2026). https://pith.science/paper/MALLLV5N

@misc{pith2026241115808,
  author       = {Pith},
  title        = {Pith review of: LRSAA: Large-scale Remote Sensing Image Target Recognition and Automatic Annotation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MALLLV5N}},
  note         = {Machine review of arXiv:2411.15808}
}
read the original abstract

This paper presents a method for object recognition and automatic labeling in large-area remote sensing images called LRSAA. The method integrates YOLOv11 and MobileNetV3-SSD object detection algorithms through ensemble learning to enhance model performance. Furthermore, it employs Poisson disk sampling segmentation techniques and the EIOU metric to optimize the training and inference processes of segmented images, followed by the integration of results. This approach not only reduces the demand for computational resources but also achieves a good balance between accuracy and speed. The source code for this project has been made publicly available on https://github.com/anaerovane/LRSAA.

Figures

Figures reproduced from arXiv: 2411.15808 by the authors.

Figure 1
Figure 1. Practical application effects of the LRSAA model: a [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 3
Figure 3. Sampling with Poisson Disk, the red points represent [PITH_FULL_IMAGE:figures/full_fig_p002_3.png] view at source ↗
Figure 2
Figure 2. Methods: Our methodology comprises four stages: Step A, B, C, and D. In Step A, we utilize Poisson disk sampling [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

23 extracted references · 19 canonical work pages

  1. [1]

    ”Support vector machines in remote sensing: A review.” ISPRS journal of photogram- metry and remote sensing 66.3 (2011): 247-259

    Mountrakis, Giorgos, Jungho Im, and Caesar Ogole. ”Support vector machines in remote sensing: A review.” ISPRS journal of photogram- metry and remote sensing 66.3 (2011): 247-259

    work page 2011

  2. [2]

    ”Ship detection from optical remote sensing images using multi-scale analysis and Fourier HOG descriptor.” Remote Sensing 11.13 (2019): 1529

    Dong, Chao, et al. ”Ship detection from optical remote sensing images using multi-scale analysis and Fourier HOG descriptor.” Remote Sensing 11.13 (2019): 1529

    work page 2019

  3. [3]

    Gui, Shengxi, Shuang Song, Rongjun Qin, and Yang Tang. 2024. ”Re- mote Sensing Object Detection in the Deep Learning Era—A Review” Remote Sensing 16, no. 2: 327. https://doi.org/10.3390/rs16020327

    doi:10.3390/rs16020327 2024

  4. [4]

    ”Attention and feature fusion SSD for remote sensing object detection.” IEEE Transactions on Instrumentation and Measurement 70 (2021): 1-9

    Lu, Xiaocong, et al. ”Attention and feature fusion SSD for remote sensing object detection.” IEEE Transactions on Instrumentation and Measurement 70 (2021): 1-9

    work page 2021

  5. [5]

    ”Mobilenets: Efficient convolutional neural net- works for mobile vision applications.” arXiv preprint arXiv:1704.04861 (2017)

    Howard, Andrew G. ”Mobilenets: Efficient convolutional neural net- works for mobile vision applications.” arXiv preprint arXiv:1704.04861 (2017)

    arXiv 2017

  6. [6]

    Howard, M

    A. Howard, M. Sandler, G. Chu, L.-C. Chen, B. Chen, M. Tan, W. Wang, Y . Zhu, R. Pang, V . Vasudevan, Q. V . Le, and H. Adam, ”Searching for MobileNetV3,” in Proc. IEEE Int. Conf. Comput. Vis. (ICCV), October 2019

    work page 2019

  7. [7]

    D. Li, A. Zhou, and A. Yao, ”HBONet: Harmonious Bottleneck on Two Orthogonal Dimensions,” in Proc. IEEE Int. Conf. Comput. Vis. (ICCV), October 2019

    work page 2019

  8. [8]

    Biswas, H

    D. Biswas, H. Su, C. Wang, A. Stevanovic, and W. Wang, ”An automatic traffic density estimation using Single Shot Detection (SSD) and MobileNet-SSD,” Physics and Chemistry of the Earth, Parts A/B/C, vol. 110, pp. 176-184, 2019

    work page 2019

Show all 23 references
  1. [9]

    Nateghi, Mohammadjavd. ”Detection, recognition and tracking cars from uav based implementation of mobilenet-single shot detection deep neural network on the embedded system by using remote sensing techniques.” Journal of Radar and Optical Remote Sensing and GIS 3.2 (2020): 53- 62

    2020

  2. [10]

    Khanam and M

    R. Khanam and M. Hussain, ”YOLOv11: An Overview of the Key Architectural Enhancements,” arXiv preprint arXiv:2410.17725, 2024

    2024 arXiv

  3. [11]

    ”You only look once: Unified, real-time object detection.” Proceedings of the IEEE conference on computer vision and pattern recognition

    Redmon, J. ”You only look once: Unified, real-time object detection.” Proceedings of the IEEE conference on computer vision and pattern recognition. 2016

    2016

  4. [12]

    Zakria, J

    Z. Zakria, J. Deng, R. Kumar, M. S. Khokhar, J. Cai and J. Kumar, ”Multiscale and Direction Target Detecting in Remote Sensing Images via Modified YOLO-v4,” in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 15, pp. 1039-1048, 2022, doi: ...

    2022

  5. [13]

    R-YOLO: A YOLO-Based Method for Arbitrary-Oriented Target Detection in High-Resolution Remote Sensing Images

    Hou Y , Shi G, Zhao Y , Wang F, Jiang X, Zhuang R, Mei Y , Ma X. R-YOLO: A YOLO-Based Method for Arbitrary-Oriented Target Detection in High-Resolution Remote Sensing Images. Sensors. 2022; 22(15):5716. https://doi.org/10.3390/s22155716

    2022 doi

  6. [14]

    J. Lin, Y . Zhao, S. Wang and Y . Tang, ”YOLO-DA: An Efficient YOLO-Based Detector for Remote Sensing Object Detection,” in IEEE Geoscience and Remote Sensing Letters, vol. 20, pp. 1-5, 2023, Art no. 6008705, doi: 10.1109/LGRS.2023.3303896

    2023

  7. [15]

    ”Stochastic sampling in computer graphics.” ACM Transactions on Graphics (TOG) 5.1 (1986): 51-72

    Cook, Robert L. ”Stochastic sampling in computer graphics.” ACM Transactions on Graphics (TOG) 5.1 (1986): 51-72

    1986

  8. [16]

    Robert L. Cook. 1986. Stochastic sampling in computer graphics. ACM Trans. Graph. 5, 1 (Jan. 1986), 51–72. https://doi.org/10.1145/7529.8927

    1986

  9. [17]

    ”A comparison of methods for generating Poisson disk distributions.” Computer Graphics Forum

    Lagae, Ares, and Philip Dutr ´e. ”A comparison of methods for generating Poisson disk distributions.” Computer Graphics Forum. V ol. 27. No. 1. Oxford, UK: Blackwell Publishing Ltd, 2008

    2008

  10. [18]

    ”Poisson-disk sampling: Theory and applications.” Ency- clopedia of Computer Graphics and Games

    Wang, Tong. ”Poisson-disk sampling: Theory and applications.” Ency- clopedia of Computer Graphics and Games. Cham: Springer Interna- tional Publishing, 2024. 1424-1431

    2024

  11. [19]

    ”Focal and efficient IOU loss for accurate bounding box regression.” Neurocomputing 506 (2022): 146-157

    Zhang, Yi-Fan, et al. ”Focal and efficient IOU loss for accurate bounding box regression.” Neurocomputing 506 (2022): 146-157

    2022

  12. [20]

    ”Distance-IoU loss: Faster and better learning for bounding box regression.” Proceedings of the AAAI conference on artificial intelligence

    Zheng, Zhaohui, et al. ”Distance-IoU loss: Faster and better learning for bounding box regression.” Proceedings of the AAAI conference on artificial intelligence. V ol. 34. No. 07. 2020

    2020

  13. [21]

    ”xview: Objects in context in overhead imagery.” arXiv preprint arXiv:1802.07856 (2018)

    Lam, Darius, et al. ”xview: Objects in context in overhead imagery.” arXiv preprint arXiv:1802.07856 (2018)

    2018 arXiv

  14. [22]

    ”Faster R-CNN: Towards real-time object detection with region proposal networks.” IEEE transactions on pattern analysis and machine intelligence 39.6 (2016): 1137-1149

    Ren, Shaoqing, et al. ”Faster R-CNN: Towards real-time object detection with region proposal networks.” IEEE transactions on pattern analysis and machine intelligence 39.6 (2016): 1137-1149

    2016

  15. [23]

    Cheng, Xun, and Jianbo Yu. ”RetinaNet with difference channel at- tention and adaptively spatial feature fusion for steel surface defect detection.” IEEE Transactions on Instrumentation and Measurement 70 (2020): 1-11. 6 APPENDIX Tianjin Shanghai Xiamen Original RetinaNet SSD ...

    2020

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.