Bounding Box Localization Quality of Modified YOLOv4-Tiny-3L in Remote Sensing Images

Penulis

  • Ahmad Fauzan Kristen Indonesia Paulus University
  • Muhammad Januansyah Jayadi Kristen Indonesia Paulus University
  • Andy Fadly Muhammadiyah Education University of Sorong

DOI:

https://doi.org/10.55123/jomlai.v5i2.9332

Kata Kunci:

Object Detection, Remote Sensing, Yolov4-Tiny, Localization Quality, Intersection Over Union, RSOD

Abstrak

Object detection in remote sensing imagery should not be evaluated only at AP50 because applications requiring accurate object positions depend on bounding-box regression quality at stricter Intersection over Union (IoU) thresholds. This study performs a secondary analysis of previously published YOLOv4-Tiny-3L and YOLOv4-Tiny-Mod experiments on the Remote Sensing Object Detection (RSOD) dataset; no model retraining was conducted. Unlike the earlier study, which focused on architectural modification and aggregate mAP, the present contribution characterizes localization degradation using AP Decay, AP Slope as a descriptive mean drop per threshold interval, and three IoU zones. YOLOv4-Tiny-Mod achieved higher AP at every evaluated threshold, with the largest gain of 13.83 percentage points at AP85; the mean advantage was 6.29 points in the moderate-precision zone and 5.66 points in strict localization. Endpoint AP Decay was nearly identical and slightly larger for the modified model (97.93 vs. 97.72), so it is not interpreted in isolation as evidence of superiority. Class-level results show a strong Recall improvement for Aircraft, whereas Oiltank and Playground retain very low Recall that cannot be causally explained from aggregate data alone. The results indicate that CSPBlock and the 104×104 detection head mainly benefit the model as localization requirements become stricter.

Referensi

[1] A. Fauzan dan N. N. R. A. Mokobombang, "Modification of YOLOv4-Tiny-3L Architecture to Improve the Accuracy of Airport Object Detection," dalam Proc. International Conference on Information Technology and Computer Science (ICTCS), 2025.

[2] X. Cheng, B. Zhang, J. Chen, dan H. Ke, "Deep Learning-Based Object Detection Techniques for Remote Sensing Images: A Survey," Remote Sensing, vol. 14, no. 10, hal. 2385, Mei 2022. [Daring]. Tersedia: https://www.mdpi.com/2072-4292/14/10/2385

[3] P. Jiang, D. Ergu, F. Liu, Y. Cai, dan B. Ma, "A Review of Yolo Algorithm Developments," Procedia Computer Science, vol. 199, hal. 1066–1073, 2022. doi: 10.1016/j.procs.2022.01.135

[4] V. K. Sharma, P. Dhiman, dan R. K. Rout, "Improved Traffic Sign Recognition Algorithm Based on YOLOv4-Tiny," Journal of Visual Communication and Image Representation, vol. 91, hal. 103774, Mar. 2023. doi: 10.1016/j.jvcir.2023.103774

[5] C.-Y. Wang, A. Bochkovskiy, dan H.-Y. M. Liao, "Scaled-YOLOv4: Scaling Cross Stage Partial Network," Feb. 2021, arXiv: arXiv:2011.08036. doi: 10.48550/arXiv.2011.08036

[6] Z. Cai dan N. Vasconcelos, "Cascade R-CNN: High Quality Object Detection and Instance Segmentation," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 43, no. 5, hal. 1483–1498, Mei 2021. doi: 10.1109/TPAMI.2019.2956516

[7] D. Cai, Z. Zhang, dan Z. Zhang, "Corner-Point and Foreground-Area IoU Loss: Better Localization of Small Objects in Bounding Box Regression," Sensors, vol. 23, no. 10, hal. 4961, Mei 2023. doi: 10.3390/s23104961

[8] S. El Ghazouali, A. Gucciardi, F. Venturini, N. Venturi, M. Rueegsegger, dan U. Michelucci, "FlightScope: An Experimental Comparative Review of Aircraft Detection Algorithms in Satellite Imagery," Remote Sensing, vol. 16, no. 24, hal. 4715, Jan. 2024. doi: 10.3390/rs16244715

[9] J. Liu dan Z. Wang, "Small Object Detection Based on Deep Learning for Remote Sensing: A Comprehensive Review," Remote Sensing, vol. 15, no. 13, hal. 3265, Jun. 2023. [Daring]. Tersedia: https://www.mdpi.com/2072-4292/15/13/3265

[10] RSIA-LIESMARS-WHU, "RSOD-Dataset," GitHub, Jan. 2025. [Daring]. Tersedia: https://github.com/RSIA-LIESMARS-WHU/RSOD-Dataset

[11] Ying-Tung Hsiao, Jia-Shing Sheu, dan Hsu Ma, "Efficient Object Detection and Intelligent Information Display Using YOLOv4-Tiny," Advances in Technology Innovation, vol. 9, no. 1, hal. 42–49, Des. 2023. doi: 10.46604/aiti.2023.12682

[12] N. T. Allo, Indrabayu, dan Z. Zainuddin, "A Novel Approach of Hybrid Bounding Box Regression Mechanism to Improve Convergency Rate and Accuracy," International Journal of Intelligent Engineering and Systems, vol. 17, no. 2, hal. 715–727, Apr. 2024. doi: 10.22266/ijies2024.0430.57

[13] Y.-H. Liao dan J.-G. Juang, "Automatic Marine Debris Inspection," Aerospace, vol. 10, no. 1, hal. 84, Jan. 2023. doi: 10.3390/aerospace10010084

[14] Z.-Q. Zhao, P. Zheng, S.-T. Xu, dan X. Wu, "Object Detection With Deep Learning: A Review," IEEE Transactions on Neural Networks and Learning Systems, vol. 30, no. 11, hal. 3212–3232, Nov. 2019. doi: 10.1109/TNNLS.2018.2876865

[15] M. L. Ali dan Z. Zhang, "The YOLO Framework: A Comprehensive Review of Evolution, Applications, and Benchmarks in Object Detection," Computers, vol. 13, no. 12, hal. 336, Des. 2024. doi: 10.3390/computers13120336

[16] N. Jegham, C. Y. Koh, M. Abdelatti, dan A. Hendawi, "Evaluating the Evolution of YOLO (You Only Look Once) Models: A Comprehensive Benchmark Study of YOLO11 and Its Predecessors," Okt. 2024, arXiv: arXiv:2411.00201. doi: 10.48550/arXiv.2411.00201

[17] C.-Y. Wang, H.-Y. M. Liao, Y.-H. Wu, P.-Y. Chen, J.-W. Hsieh, dan I.-H. Yeh, "CSPNet: A New Backbone That Can Enhance Learning Capability of CNN," dalam Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2020, hal. 390–391.

[18] H. Rezatofighi, N. Tsoi, J. Gwak, A. Sadeghian, I. Reid, dan S. Savarese, "Generalized Intersection Over Union: A Metric and a Loss for Bounding Box Regression," dalam Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019, hal. 658–666.

[19] Z. Zheng, P. Wang, W. Liu, J. Li, R. Ye, dan D. Ren, "Distance-IoU Loss: Faster and Better Learning for Bounding Box Regression," Proceedings of the AAAI Conference on Artificial Intelligence, vol. 34, no. 07, hal. 12993–13000, 2020. doi: 10.1609/aaai.v34i07.6999.

[20] T.-Y. Lin, P. Goyal, R. Girshick, K. He, dan P. Dollár, "Focal Loss for Dense Object Detection," dalam Proc. IEEE International Conference on Computer Vision (ICCV), 2017, hal. 2980–2988.

Unduhan

Diterbitkan

2026-06-15

Cara Mengutip

Ahmad Fauzan, Muhammad Januansyah Jayadi, & Andy Fadly. (2026). Bounding Box Localization Quality of Modified YOLOv4-Tiny-3L in Remote Sensing Images. JOMLAI: Journal of Machine Learning and Artificial Intelligence, 5(2), 104–111. https://doi.org/10.55123/jomlai.v5i2.9332

Terbitan

Vol 5 No 2 (2026): Juni 2026

Bagian

Articles

Lisensi

Hak Cipta (c) 2026 Ahmad Fauzan, Muhammad Januansyah Jayadi, Andy Fadly

Creative Commons License

Artikel ini berlisensiCreative Commons Attribution-ShareAlike 4.0 International License.

Copyright (c) 2022 The authors. Published by Yayasan Literasi Indonesia
Creative Commons License This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License

The author(s) whose article is published in the JOMLAI journal attain the copyright for their article and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution-ShareAlike 4.0 International License. By submitting the manuscript to JOMLAI, the author(s) agree with this policy. No special document approval is required.

The author(s) guarantee that:their article is original, written by the mentioned author(s),

  • has never been published before,
  • does not contain statements that violate the law, and
  • does not violate the rights of others, is subject to copyright held exclusively by the author(s), and is free from the rights of third parties, and that the necessary written permission to quote from other sources has been obtained by the author(s).

The author(s) retain all rights to the published work, such as (but not limited to) the following rights:

  • Copyright and other proprietary rights related to the article, such as patents,
  • The right to use the substance of the article in its own future works, including lectures and books,
  • The right to reproduce the article for its own purposes,
  • The right to archive all versions of the article in any repository, and
  • The right to enter into separate additional contractual arrangements for the non-exclusive distribution of published versions of the article (for example, posting them to institutional repositories or publishing them in a book), acknowledging its initial publication in this journal (JOMLAI: Journal of Machine Learning and Artificial Intelligence).

Suppose the article was prepared jointly by more than one author. Each author submitting the manuscript warrants that all co-authors have given their permission to agree to copyright and license notices (agreements) on their behalf and notify co-authors of the terms of this policy. JOMLAI will not be held responsible for anything that may arise because of the writer's internal dispute. JOMLAI will only communicate with correspondence authors.

Authors should also understand that their articles (and any additional files, including data sets, and analysis/computation data) will become publicly available once published. The license of published articles (and additional data) will be governed by a Creative Commons Attribution-ShareAlike 4.0 International License. JOMLAI allows users to copy, distribute, display and perform work under license. Users need to attribute the author(s) and JOMLAI to distribute works in journals and other publication media. Unless otherwise stated, the author(s) is a public entity as soon as the article is published