Skripsi

DETEKSI KERUSAKAN JALAN MENGGUNAKAN YOLOV11 DAN SLICING AIDED HYPER INFERENCE

---

This study investigates automated road damage detection by combining the YOLOv11 algorithm with the Slicing Aided Hyper Inference (SAHI) approach on the multinational RDD2022 dataset, covering four damage categories: Longitudinal Crack (D00), Transverse Crack (D10), Alligator Crack (D20), and Pothole (D40). Ten YOLOv11 training experiments and eleven SAHI hyperparameter configurations were conducted to identify optimal settings. The YOLOv11s model at 768-pixel input resolution (EXP-5) was established as the best baseline configuration with mAP@50 = 0.6534, Precision = 0.6532, Recall = 0.6058, and an inference time of 21.62 ms per image (46.24 FPS). Applying the optimal SAHI configuration (SAHI-09: Slice Size 896px, Overlap Ratio 0.20, Match Threshold 0.40) improved Precision by +2.87% to 0.6719, with an inference time of 30.90 ms per image (32.4 FPS) that remains above the 30 FPS real-time processing threshold. Notably, the D40 Pothole category demonstrated a Recall improvement of +1.66%, confirming SAHI's effectiveness in detecting small and irregular objects. This study provides empirical evidence that the integration of YOLOv11 and SAHI yields an accurate road damage inspection system that remains operationally viable for real-time deployment.

---

Availability

Detail Information

Series Title

-

Call Number

T2008822026

Publisher

Indralaya : Prodi Teknik Informatika, Fakultas Ilmu Komputer Universitas Sriwijaya., 2026

Collation

xvi, 148 hlm.; ilus.; tab.; 29 cm.

Language

Indonesia

ISBN/ISSN

-

Classification

625.760 7

Content Type

Text

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

teknik jalan raya
Prodi Teknik Informatika

Specific Detail Info

-

Statement of Responsibility

MI

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment