Pith. sign in

REVIEW 4 major objections 3 minor 34 references

Synthetic Dataset Generation for Autonomous Mobile Robots Using 3D Gaussian Splatting for Vision Training

T0 review · 4 major / 3 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read 3D Gaussian splats can generate visual training data for robots nearly as well as real photos.

desk verdict A useful sim-to-real pipeline paper whose central hybrid-data claim is undercut by a dataset-size confound; the ball experiments are solid, the robot experiments need a same-size baseline. read the letter →

arxiv 2506.05092 v1 pith:CSKE5JOT submitted 2025-06-05 cs.RO cs.CV

Aneesh Deogan , Wout Beks , Peter Teurlings , Koen de Vos , Mark van den Brand , Rene van de Molengraft This is my paper
classification cs.ROcs.CV
keywords syntheticdatageneration3DGaussiansplattingobjectdetectionrobotsoccerYOLOv8domaintransferautonomousmobilerobotsautomatedannotation
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 photorealistic 3D Gaussian splats of real objects, placed in a virtual soccer arena, can replace most hand-annotated real images when training an object detector. In their experiments, a YOLOv8 detector trained only on synthetic images reaches a mean average precision of 0.990 for the ball, against 0.994 for a detector trained on real photos; for robot detection, adding synthetic images to a real dataset (0.992 mAP50) beats using either alone (0.973 real-only, 0.942 synthetic-only). Because the simulator knows every object coordinate and camera pose, labels are generated automatically and almost instantly, so the pipeline is far cheaper than manual annotation. The authors present this as the first synthetic-training demonstration for robot soccer and as a scalable route to diverse annotated data in dynamic robotics.

What carries the argument

The load-bearing object is the 3D Gaussian splat: a rasterization technique that learns a scene as a set of depth-sorted Gaussian ellipsoids, optimized by stochastic gradient descent from a small set of images, and renders photorealistically in real time. The authors use these splats for the target objects, arrange them in a virtual field that mimics the real arena's lighting and the onboard camera's field of view, and exploit the simulator's exact knowledge of geometry to produce bounding-box labels automatically. The detector is YOLOv8, kept identical across all datasets, so differences in validation metrics isolate the effect of data source.

What would settle it

Take a held-out set of real match images from a different arena or with a different onboard camera and train a detector using the paper's synthetic generation pipeline; if the mAP50 gap relative to a real-data-trained detector exceeds the roughly 0.4-to-3-point differences reported here, the domain-fidelity premise is falsified. A second check: render synthetic test images from the same 3DGS models and confirm whether the reported parity persists on real images but drops significantly on synthetic renderings, which would indicate the detector is fitting generic texture features rather than generalizable object geometry.

Watch

Extended reading notes

Core claim

The central discovery is that 3D Gaussian splatting provides the missing piece for synthetic data generation in robotics: photorealistic object models that render in real time. Using smartphone photos processed by a commercial reconstruction service, the authors build splats of a ball and three soccer robots, import them into a game-engine simulation of a soccer field, randomize lighting and camera pose, and automatically project known 3D coordinates into 2D bounding boxes. Validation on real match footage shows a synthetic-only detector reaching mAP50 of 0.929 over four classes, and a combined real-plus-synthetic detector outperforming a real-only detector for robot localization (0.992 vs 0.973). The paper concludes that the method is a practical, scalable alternative to manual dataset creation for high-dynamic robotic environments.

Load-bearing premise

The premise is that the reconstructed object models and the virtual arena reproduce the real match conditions—lighting, camera field of view, object appearance—closely enough that a detector trained purely on synthetic frames transfers to real footage; the paper asserts this fidelity but does not quantitatively verify it against real image statistics.

Editorial extensions

If this is right

  • A team with a phone and a laptop can generate tens of thousands of annotated training images in hours, enabling rapid detector deployment for new objects or altered robot appearances.
  • The hybrid training recipe, real images for quality and synthetic images for scale, becomes a general template for scarce-data robotics domains beyond soccer.
  • Because pose, lighting, and background are all controllable in the simulator, the same pipeline can deliberately oversample rare or safety-critical scenarios such as close encounters or crowded fields.
  • An annotation cost of about one second per image, versus seconds plus collection time for real data, makes continuous retraining on fresh synthetic data practical without human labelling effort.

Reading between the lines

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

  • The paper's parity claim is validated only on images from the same arena and camera configuration; a natural stress test is to hold out a different arena, lighting season, or camera lens and measure how much the mAP advantage shrinks.
  • The reported high precision and lower recall suggest that an explicit false-positive penalty during training, aligned with the authors' stated preference for avoiding costly misidentifications, could push synthetic-only performance closer to or beyond the real-data baseline.
  • Since the 3DGS models build in about an hour from a handful of images, the pipeline could be reused for any robot manipulation task, effectively bootstrapping a detector for a novel object without any manual labeling.
  • A controlled comparison between 3DGS-based rendering and non-photorealistic simulated objects of the same robots would separate the contribution of photorealism from that of scene diversity in the transfer 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

4 major / 3 minor

Summary. This paper presents a pipeline for generating synthetic annotated training data for object detection by reconstructing 3D Gaussian splats of target objects (a ball and soccer robots) from smartphone images and placing them in an Unreal Engine virtual environment with configurable lighting, background, and domain randomization. The authors train YOLOv8 detectors on real, synthetic, and combined datasets and report mAP/F1 on held-out real images in three settings: ball detection (Table I), robot detection (Table II), and multi-class match gameplay (Table III). The main claims are that 3DGS synthetic data reaches near real-data performance much faster than manual annotation, and that combining real and synthetic data significantly improves detection.

Significance. If the claims hold, the method offers a practical, low-cost alternative for building annotated datasets in dynamic robotics domains: annotations are generated automatically by projection, the pipeline is described in enough detail to reproduce, and the evaluation uses held-out real images, so there is no circularity in the main comparisons. The time-to-dataset comparisons (1 second per synthetic image versus 8 hours of manual annotation for the robot set) are useful for practitioners. However, the paper's headline claim about hybrid training is currently supported only by a confounded comparison, and the statistical basis is thin because validation sets are small and no repeated runs are reported. The significance is therefore conditional on strengthening the experimental design.

major comments (4)
  1. [Section IV-B, Table II] The central evidence for the claim that combining real and synthetic data 'significantly increases' performance is confounded by training-set size. Combined-6000 (2000 real + 4000 synthetic) is compared only against Real-World-2300, so the gain from 0.973 to 0.992 mAP50 can be explained by the larger number of training images rather than by a real+synthetic synergy. The progression 3DGS-2300 (0.813) to 3DGS-6000 (0.942) to Combined-6000 (0.992) is consistent with a pure dataset-size effect. The authors should add Real-World-6000 and, ideally, a 2000-real + 4000-synthetic split with a matching real-only 6000-image baseline, and show that the combined model outperforms both at equal size.
  2. [Section IV-A/IV-B and Abstract] The word 'significantly' in the abstract is not supported by the reported experiments: validation uses only 70 images in Tables I-II and 85 images in Table III, each experiment is run once, and no error bars, confidence intervals, or significance tests are reported. The authors should either provide repeated runs with variance estimates and a statistical test (e.g., bootstrap over validation images or multiple training seeds), or replace 'significantly' with a weaker claim such as 'can improve.'
  3. [Section III-A/III-B and IV-C] The faithfulness of the 3DGS models and the simulation environment (lighting, FOV, background) to real match conditions is asserted but not quantitatively verified. Because the entire method rests on the transferability of synthetic images to real footage, the paper should include at least a domain-gap analysis (e.g., distribution shift in image statistics, occlusion/pose coverage) or an ablation showing the effect of the claimed alignment choices (ZED2 FOV replication, lighting mimicry, background enhancement) on detection performance.
  4. [Section IV-C, Table III] In the match-play validation, the 3DGS-trained model is not compared with a real-world-trained baseline, so the claim of effectiveness in multi-class scenarios is only a single-arm demonstration. The authors acknowledge the lack of a similarly sized real dataset, but they should temper the conclusion or compare against a real-data-only model trained on whatever real data exists for the four classes, even at smaller size.
minor comments (3)
  1. [Throughout] Typographical issues include 'Koen de V os', 'R ESULTS', and the non-standard possessive '3DGS's'; a proofreading pass is needed.
  2. [Section III-C] The sentence 'All trainings are executed...' and the phrase 'approximately 1000 images without Automatic Mixed Precision (AMP) training' are ambiguous; specify the exact number of images per ball dataset and clarify whether AMP is disabled for all trainings.
  3. [Section IV-A] The phrase 'the mAP50 difference of 0.4%' refers to 0.990 versus 0.994; this is 0.4 percentage points, not 0.4% relative. Please use consistent notation for percentage-point differences.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: synthetic training and real validation are independent, held-out data sources.

full rationale

The paper's derivation chain is: (1) capture smartphone images of the ball and robots; (2) reconstruct 3DGS models with LUMA AI; (3) render synthetic images in Unreal Engine with automated bounding boxes from known 3D coordinates; (4) train YOLOv8 on synthetic, real, or combined datasets; (5) evaluate on hand-annotated real-world images (70 images for ball/robot tests, 85 for match gameplay). The evaluation set is held out and manually labelled, and the synthetic training images are not derived from those validation images, so the reported mAP values are not equivalent by construction to any fitted input. No load-bearing self-citation appears: the cited 3DGS method [21] and YOLO [31] are external tools, and the paper does not invoke a prior 'uniqueness theorem' or define its target metric in terms of its model. The paper explicitly acknowledges limitations in Section IV-C (goalkeeper robots omitted from training and therefore undetected; synthetic data contains no motion blur), and Section IV-B admits that real data outperforms 3DGS at equal dataset size; these are honest scope statements, not circular steps. The main methodological weakness is a confound in Section IV-B/Table II: Combined-6000 (2000 real + 4000 synthetic) is compared only with Real-World-2300, not a 6000-image real dataset, so the claimed 'significantly increases' hybrid benefit may be a dataset-size effect; but a confounded comparison is a correctness/validity issue, not a reduction of the conclusion to its inputs. No circular step can be quoted, so the score is 0.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The paper introduces no new physical or conceptual entities. It relies on existing methods (3DGS, Unreal, YOLO) and hand-chosen simulation parameters. The main unverified inputs are the fidelity assumptions of the synthetic pipeline and the accuracy of the small validation set.

free parameters (1)
  • Domain randomization ranges (light intensity, camera orientation, ball/robot orientation) = Not quantified in paper
    The paper states these variations are application-specific and uses uniform distributions within predefined bounds, but exact ranges are not reported. If these were tuned to maximize validation performance, the central result could be partly fitted; no sensitivity analysis is provided.
assumptions (3)
  • domain assumption 3D Gaussian splat models are sufficiently photorealistic for sim-to-real transfer.
    The entire method depends on 3DGS models reducing the domain gap; Section III-A describes model creation but provides no quantitative fidelity check.
  • domain assumption The Unreal Engine camera accurately replicates the ZED2 lens parameters and FOV.
    Section III-B states the camera replicates lens parameters and FOV, but no calibration or comparison is shown, so an inaccurate replica would degrade transfer.
  • domain assumption The hand-annotated real-world validation images are accurate ground truth.
    All reported metrics rely on manual annotations of 70-85 real images; annotation errors would directly bias the comparisons.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Synthetic Dataset Generation for Autonomous Mobile Robots Using 3D Gaussian Splatting for Vision Training." pith.science (2026). https://pith.science/paper/CSKE5JOT

@misc{pith2026250605092,
  author       = {Pith},
  title        = {Pith review of: Synthetic Dataset Generation for Autonomous Mobile Robots Using 3D Gaussian Splatting for Vision Training},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CSKE5JOT}},
  note         = {Machine review of arXiv:2506.05092}
}
read the original abstract

Annotated datasets are critical for training neural networks for object detection, yet their manual creation is time- and labour-intensive, subjective to human error, and often limited in diversity. This challenge is particularly pronounced in the domain of robotics, where diverse and dynamic scenarios further complicate the creation of representative datasets. To address this, we propose a novel method for automatically generating annotated synthetic data in Unreal Engine. Our approach leverages photorealistic 3D Gaussian splats for rapid synthetic data generation. We demonstrate that synthetic datasets can achieve performance comparable to that of real-world datasets while significantly reducing the time required to generate and annotate data. Additionally, combining real-world and synthetic data significantly increases object detection performance by leveraging the quality of real-world images with the easier scalability of synthetic data. To our knowledge, this is the first application of synthetic data for training object detection algorithms in the highly dynamic and varied environment of robot soccer. Validation experiments reveal that a detector trained on synthetic images performs on par with one trained on manually annotated real-world images when tested on robot soccer match scenarios. Our method offers a scalable and comprehensive alternative to traditional dataset creation, eliminating the labour-intensive error-prone manual annotation process. By generating datasets in a simulator where all elements are intrinsically known, we ensure accurate annotations while significantly reducing manual effort, which makes it particularly valuable for robotics applications requiring diverse and scalable training data.

Figures

Figures reproduced from arXiv: 2506.05092 by the authors.

Figure 1
Figure 1. The proposed pipeline for generating synthetic datasets for object detection using 3D Gaussian splats. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The (a) simulated and (b) 3D Gaussian splat balls used [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 4
Figure 4. Example of the virtual environment, including the [PITH_FULL_IMAGE:figures/full_fig_p003_4.png] view at source ↗
Figures from the paper (1 more)
Figure 5
Figure 5. Figure 5: Validation on real life match gameplay. (Ball: b, Tech [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

34 extracted references · 26 canonical work pages

  1. [1]

    Robotics and industry 4.0,

    R. Goel and P. Gupta, “Robotics and industry 4.0,” A Roadmap to Industry 4.0: Smart Production, Sharp Business and Sustainable Development, pp. 157–169, 2020

    work page 2020

  2. [2]

    3d object detection and tracking methods using deep learning for computer vision applications,

    E. Shreyas, M. H. Sheth et al. , “3d object detection and tracking methods using deep learning for computer vision applications,” in 2021 International Conference on Recent Trends on Electronics, Information, Communication & Technology (RTEICT) . IEEE, 2021, pp. 735–738

    work page 2021

  3. [3]

    Deep learning vs. traditional computer vision,

    N. O’Mahony, S. Campbell, A. Carvalho, S. Harapanahalli, G. V . Hernandez, L. Krpalkova, D. Riordan, and J. Walsh, “Deep learning vs. traditional computer vision,” in Advances in Computer Vision: Proceedings of the 2019 Computer Vision Conference (CVC), Volume 1 1. Springer, 2020, pp. 128–144

    work page 2019

  4. [4]

    Machine learning and deep learning,

    C. Janiesch, P. Zschech, and K. Heinrich, “Machine learning and deep learning,” Electronic Markets, vol. 31, no. 3, pp. 685–695, 2021

    work page 2021

  5. [5]

    You only look once: Unified, real-time object detection,

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

    work page 2016

  6. [6]

    A review of object detection models based on convolutional neural network,

    F. Sultana, A. Sufian, and P. Dutta, “A review of object detection models based on convolutional neural network,” Intelligent computing: image processing based applications , pp. 1–16, 2020

    work page 2020

  7. [7]

    Revisiting unrea- sonable effectiveness of data in deep learning era,

    C. Sun, A. Shrivastava, S. Singh, and A. Gupta, “Revisiting unrea- sonable effectiveness of data in deep learning era,” in Proceedings of the IEEE international conference on computer vision , 2017, pp. 843–852

    work page 2017

  8. [8]

    Object recognition datasets and challenges: A review,

    A. Salari, A. Djavadifar, X. Liu, and H. Najjaran, “Object recognition datasets and challenges: A review,” Neurocomputing, vol. 495, pp. 129–152, 2022

    work page 2022

Show all 34 references
  1. [9]

    Microsoft coco: Common objects in context,

    T.-Y . Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Doll ´ar, and C. L. Zitnick, “Microsoft coco: Common objects in context,” in Computer Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13 . Springer,...

    2014

  2. [10]

    Analyzing computer vision data-the good, the bad and the ugly,

    O. Zendel, K. Honauer, M. Murschitz, M. Humenberger, and G. Fer- nandez Dominguez, “Analyzing computer vision data-the good, the bad and the ugly,” in Proceedings of the IEEE Conference on Com- puter Vision and Pattern Recognition , 2017, pp. 1980–1990

    2017

  3. [11]

    Image-bot: Generating synthetic object detection datasets for small and medium- sized manufacturing companies,

    L. Block, A. Raiser, L. Sch ¨on, F. Braun, and O. Riedel, “Image-bot: Generating synthetic object detection datasets for small and medium- sized manufacturing companies,” Procedia CIRP, vol. 107, pp. 434– 439, 2022

    2022

  4. [12]

    Review and analysis of synthetic dataset generation methods and techniques for application in computer vision,

    G. Paulin and M. Ivasic-Kos, “Review and analysis of synthetic dataset generation methods and techniques for application in computer vision,” Artificial intelligence review, vol. 56, no. 9, pp. 9221–9265, 2023

    2023

  5. [13]

    Ovvv: Using virtual worlds to design and evaluate surveillance systems,

    G. R. Taylor, A. J. Chosak, and P. C. Brewer, “Ovvv: Using virtual worlds to design and evaluate surveillance systems,” in 2007 IEEE conference on computer vision and pattern recognition . IEEE, 2007, pp. 1–8

    2007

  6. [14]

    Self- supervised data generation for precision agriculture: Blending sim- ulated environments with real imagery,

    L. Saraceni, I. M. Motoi, D. Nardi, and T. A. Ciarfuglia, “Self- supervised data generation for precision agriculture: Blending sim- ulated environments with real imagery,” in 2024 IEEE 20th Interna- tional Conference on Automation Science and Engineering (CASE) . IEEE, 2024, ...

    2024

  7. [15]

    Synthetic datasets for autonomous driving: A survey,

    Z. Song, Z. He, X. Li, Q. Ma, R. Ming, Z. Mao, H. Pei, L. Peng, J. Hu, D. Yao et al. , “Synthetic datasets for autonomous driving: A survey,” IEEE Transactions on Intelligent Vehicles , 2023

    2023

  8. [16]

    Dyvir: dynamic virtual reality dataset for aerial threat object detection,

    G. Williams, G. D. Lecakes Jr, A. Almon, N. Koutsoubis, K. Naddeo, T. Kiel, G. Ditzler, and N. C. Bouaynaya, “Dyvir: dynamic virtual reality dataset for aerial threat object detection,” in Synthetic Data for Artificial Intelligence and Machine Learning: Tools, Techniques, and ...

    2023

  9. [17]

    On rendering synthetic images for training an object detector,

    A. Rozantsev, V . Lepetit, and P. Fua, “On rendering synthetic images for training an object detector,” Computer Vision and Image Under- standing, vol. 137, pp. 24–37, 2015

    2015

  10. [18]

    Google scanned objects: A high-quality dataset of 3d scanned household items,

    L. Downs, A. Francis, N. Koenig, B. Kinman, R. Hickman, K. Rey- mann, T. B. McHugh, and V . Vanhoucke, “Google scanned objects: A high-quality dataset of 3d scanned household items,” in 2022 International Conference on Robotics and Automation (ICRA). IEEE, 2022, pp. 2553–2560

    2022

  11. [19]

    Generative adversarial nets,

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y . Bengio, “Generative adversarial nets,” Advances in neural information processing systems , vol. 27, 2014

    2014

  12. [20]

    Nerf: Representing scenes as neural radiance fields for view synthesis,

    B. Mildenhall, P. P. Srinivasan, M. Tancik, J. T. Barron, R. Ramamoor- thi, and R. Ng, “Nerf: Representing scenes as neural radiance fields for view synthesis,” Communications of the ACM , vol. 65, no. 1, pp. 99–106, 2021

    2021

  13. [21]

    3d gaussian splatting for real-time radiance field rendering

    B. Kerbl, G. Kopanas, T. Leimk ¨uhler, and G. Drettakis, “3d gaussian splatting for real-time radiance field rendering.” ACM Trans. Graph., vol. 42, no. 4, pp. 139–1, 2023

    2023

  14. [22]

    How much real data do we actually need: Analyzing object detection performance using synthetic and real data,

    F. E. Nowruzi, P. Kapoor, D. Kolhatkar, F. A. Hassanat, R. La- ganiere, and J. Rebut, “How much real data do we actually need: Analyzing object detection performance using synthetic and real data,” arXiv:1907.07061, 2019

    1907 arXiv

  15. [23]

    Bakedsdf: Meshing neural sdfs for real-time view synthesis,

    L. Yariv, P. Hedman, C. Reiser, D. Verbin, P. P. Srinivasan, R. Szeliski, J. T. Barron, and B. Mildenhall, “Bakedsdf: Meshing neural sdfs for real-time view synthesis,” in ACM SIGGRAPH 2023 Conference Proceedings, 2023, pp. 1–9

    2023

  16. [24]

    Realistic surgical image dataset generation based on 3d gaussian splatting,

    T. Zeng, G. Loza Galindo, J. Hu, P. Valdastri, and D. Jones, “Realistic surgical image dataset generation based on 3d gaussian splatting,” in International Conference on Medical Image Computing and Computer-Assisted Intervention . Springer, 2024, pp. 510–519

    2024

  17. [25]

    RoboCup robocup federation official website,

    “RoboCup robocup federation official website,” https://www.robocup. org/, accessed: 2025-02-24

    2025

  18. [26]

    Structure-from-motion revisited,

    J. L. Schonberger and J.-M. Frahm, “Structure-from-motion revisited,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 4104–4113

    2016

  19. [27]

    Gaussian splats for 3d visualization,

    “Gaussian splats for 3d visualization,” accessed: 2025-02-24. [Online]. Available: https://lumalabs.ai/

    2025

  20. [28]

    Tech united eindhoven,

    “Tech united eindhoven,” https://www.techunited.nl/, accessed: 2025- 02-24

    2025

  21. [29]

    Falcons robocup msl,

    “Falcons robocup msl,” https://www.falcons-robocup.nl/, accessed: 2025-02-24

    2025

  22. [30]

    2d bounding box for deep learning image detection,

    “2d bounding box for deep learning image detection,” accessed: 2025-02-24. [Online]. Available: https://www.fab.com/ listings/46a82ff2-8e9b-4a04-9434-7c7c8bebac63

    2025

  23. [31]

    Ultralytics YOLO,

    G. Jocher, J. Qiu, and A. Chaurasia, “Ultralytics YOLO,” Jan. 2023. [Online]. Available: https://github.com/ultralytics/ultralytics

    2023

  24. [32]

    A survey on performance metrics for object-detection algorithms,

    R. Padilla, S. L. Netto, and E. A. Da Silva, “A survey on performance metrics for object-detection algorithms,” in 2020 international con- ference on systems, signals and image processing (IWSSIP) . IEEE, 2020, pp. 237–242

    2020

  25. [33]

    The real- time object detection system on mobile soccer robot using yolo v3,

    H. Soebhakti, S. Prayoga, R. A. Fatekha, and M. B. Fashla, “The real- time object detection system on mobile soccer robot using yolo v3,” in 2019 2nd International Conference on Applied Engineering (ICAE) . IEEE, 2019, pp. 1–6

    2019

  26. [34]

    labelImg,

    Tzutalin, “labelImg,” Feb. 2018. [Online]. Available: https://github. com/HumanSignal/labelImg

    2018

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.