Skripsi

DESAIN APLIKASI DETEKSI DAN ESTIMASI DIMENSI LUBANG BERBASIS FLUTTER DAN DEEP LEARNING

---

Road damage, particularly potholes, can cause discomfort and increase the risk of accidents. Rapid and accurate identification of road conditions is essential for immediate repairs. However, pothole identification is still performed manually, necessitating the automation of pothole detection using digital image processing and deep learning to detect and estimate pothole dimension. This research aims to design a pothole detection and dimension estimation application based on Flutter and deep learning, namely POTION AI. The application uses YOLOv8 and Mask R-CNN algorithms to detect potholes and measure their dimensions. Training data for the model were collected using a GoPro Hero 8 camera in Sumatra Selatan, including Jl. Ariodillah, Jl. Kaca Piring, Jl. Swakarya I, Jl. Swakarya II, Jl. Dwikora II, and Jl. Kampung Bali. The training process was conducted on Google Colaboratory using YOLOv8x-seg, the largest model of YOLOv8, with 71 million parameters. Research results show that the YOLOv8 and Mask R-CNN models can detect potholes with high accuracy, achieving a confidence score above 92.22%, and performing consistently well on both local systems and mobile applications. The application testing was carried out by integrating both models, YOLOv8 and Mask R-CNN, into a Flutter-based application to detect and estimate pothole dimensions. The application also uses Leaflet JS to display an interactive map showing the detected pothole locations. Testing results indicate that the POTION AI application functions well on various devices and provides accurate information about road conditions, achieving a final usability score of 4.8875. This application is expected to help expedite road repairs and reduce accidents caused by potholes.

---

Availability

Detail Information

Series Title

-

Call Number

T1514912024

Publisher

Indralaya : Prodi Teknik Elektro, Fakultas Teknik Universitas Sriwijaya., 2024

Collation

xvi, 107 hlm.; ilus.; 29 cm

Language

Indonesia

ISBN/ISSN

-

Classification

005.250 7

Content Type

Text

Media Type

unmediated

Carrier Type

-

Edition

-

Subject(s)

Prodi Teknik Elektro
Mask R-CNN

Specific Detail Info

-

Statement of Responsibility

KA

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment