GeoSeverity-RDD: Framework Deteksi dan Estimasi Keparahan Kerusakan Jalan Indonesia Berbasis Multi-Task YOLOv11 dengan Proxy Severity Geometris
Keywords:
GeoSeverity-RDD, Proxy Severity Geometris, Road Damage Detection, Severity Estimation, YOLOv11Abstract
Penelitian ini mengusulkan GeoSeverity-RDD, kerangka kerja multi-tugas yang secara simultan mendeteksi jenis kerusakan jalan dan mengestimasi tingkat keparahannya (ringan/sedang/berat) dalam satu inferensi tunggal pada dataset jalan Indonesia. Novelty utama terletak pada Proxy Severity Geometris (PSG) yaitu mekanisme pemberian label severity otomatis yang diturunkan dari fitur geometri bounding box berupa area untuk pothole dan panjang dominan untuk cracking tanpa memerlukan anotasi manual tambahan. Dataset Road Damage Indonesia v5 (format COCO) yang digunakan terdiri dari 3.321 citra beresolusi 640×640 piksel dengan 6.999 anotasi mencakup empat kelas yaitu pothole, longitudinal cracking, lateral cracking, dan alligator cracking. Setelah penerapan PSG dengan ambang batas persentil P33/P67, distribusi label severity mencapai keseimbangan 33% per level yang berhasil mengeliminasi class imbalance. Arsitektur GeoSeverity-RDD diimplementasikan dengan menambahkan Severity Estimation Head (SEH) ringan secara paralel pada detection head YOLOv11m, serta berbagi backbone C3k2+SPPF dan neck C2PSA+FPN/PAN. Hasil eksperimen menunjukkan model mencapai [email protected] sebesar 71,3% untuk detection task dengan degradasi minimal 1,1% dari baseline. Pada severity task, model menghasilkan akurasi sebesar 78,6% dan Macro F1 78,2% dengan kecepatan inferensi 34,7 FPS (real-time). Integrasi model ini efisien dengan overhead parameter hanya meningkat 6,5% dan penambahan ukuran model sebesar 2,8 MB
References
[1] M. S. Fuady et al., “THE INFLUENCE OF INFRASTRUCTURE EXPENDITURES FROM THE MINISTRY OF PUBLIC WORKS AND PUBLIC HOUSING ON THE CONDITION OF NATIONAL ROADS BY PROVINCE 2020-2021,” 2024.
[2] N. Tanan, A. Sjafruddin, and M. Idris, “The Characteristics of Road Crashes on Indonesian National Roads Based on Integrated Road Safety Management System Data,” Periodica Polytechnica Transportation Engineering, vol. 53, no. 1, pp. 67–76, 2025, doi: 10.3311/PPtr.36654.
[3] Z. Zhong and Z. Ren, “Deep learning and TOPSIS-based multi-criteria decision-making framework for urban road defect detection and sustainable maintenance planning,” Sci. Rep., vol. 16, Dec. 2025, doi: 10.1038/s41598-025-31682-y.
[4] H. Maeda, Y. Sekimoto, T. Seto, T. Kashiyama, and H. Omata, “Road Damage Detection and Classification Using Deep Neural Networks with Smartphone Images,” Computer-Aided Civil and Infrastructure Engineering, vol. 33, no. 12, pp. 1127–1141, Dec. 2018, doi: 10.1111/mice.12387.
[5] D. Arya, H. Maeda, S. K. Ghosh, D. Toshniwal, and Y. Sekimoto, “RDD2022: A multi-national image dataset for automatic Road Damage Detection,” 2022. [Online]. Available: https://arxiv.org/abs/2209.08538
[6] Y. Indrihapsari, D. Wijaya, S. Adhiyaksa, I. Siswanto, D. Ardiansyah, and W. Ardianto, “Optimizing YOLO Models for Enhanced Road Damage Detection: A Performance Comparison of YOLOv5 and YOLOv8,” Elinvo (Electronics, Informatics, and Vocational Education), vol. 10, pp. 147–168, Jun. 2026, doi: 10.21831/elinvo.v10i2.88919.
[7] G. Jocher, J. Qiu, M. Liu, S. Lyu, F. C. Akyon, and M. E. Kalfaoglu, “Ultralytics YOLO26: Unified Real-Time End-to-End Vision Models,” 2026, doi: 10.48550/arXiv.2606.03748.
[8] T.-Y. Lin, P. Dollár, R. Girshick, K. He, B. Hariharan, and S. Belongie, “Feature Pyramid Networks for Object Detection,” 2017. [Online]. Available: https://arxiv.org/abs/1612.03144
[9] K. He, X. Zhang, S. Ren, and J. Sun, “Deep Residual Learning for Image Recognition,” 2015. [Online]. Available: https://arxiv.org/abs/1512.03385
[10] T. S. Tran, V. P. Tran, H. J. Lee, J. M. Flores, and V. P. Le, “A two-step sequential automated crack detection and severity classification process for asphalt pavements,” International Journal of Pavement Engineering, vol. 23, no. 6, pp. 2019–2033, 2022, doi: 10.1080/10298436.2020.1836561.
[11] ASTM International, Standard Practice for Roads and Parking Lots Pavement Condition Index Surveys (ASTM D6433-18). West Conshohocken, PA, 2018.
[12] N. J. Owor, H. Du, A. Daud, A. Aboah, and Y. Adu-Gyamfi, “Image2PCI – A Multitask Learning Framework for Estimating Pavement Condition Indices Directly from Images,” 2023. [Online]. Available: https://arxiv.org/abs/2310.08538
[13] Roboflow, “Road Damage Indonesia Dataset,” Oct. 2023, Roboflow. [Online]. Available: https://universe.roboflow.com/kantor-uskvs/road-damage-indonesia
[14] T.-Y. Lin et al., “Microsoft COCO: Common Objects in Context,” 2015. [Online]. Available: https://arxiv.org/abs/1405.0312
[15] L. li and K. Wang, “Bounding Box–Based Technique for Pavement Crack Classification and Measurement Using 1 mm 3D Laser Data,” Journal of Computing in Civil Engineering, vol. 30, p. 4016011, Jul. 2016, doi: 10.1061/(ASCE)CP.1943-5487.0000568.
[16] Y. Hu, R. Rayhana, L. Bai, and Z. Liu, “Computational intelligence for road pavement condition assessment: a deep learning perspective,” Urban Lifeline, vol. 4, no. 1, p. 18, 2026, doi: 10.1007/s44285-026-00073-8.
[17] G. Ghiasi et al., “Simple Copy-Paste is a Strong Data Augmentation Method for Instance Segmentation,” 2021. [Online]. Available: https://arxiv.org/abs/2012.07177
Downloads
Published
How to Cite
Issue
Section
License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
