← All projects

Road Condition Classifier & Pothole Early-Warning System

January 15, 2025 PythonTensorFlowCNNKotlinAndroidGIS

Overview

This project implements a Convolutional Neural Network (CNN) to automatically classify road conditions (good, light damage, heavy damage / potholes) from images, paired with an Android-based Geographic Information System that maps damaged segments and issues early warnings for potholes. It was my undergraduate thesis at Universitas Negeri Medan and was published in the Indonesian Journal of Artificial Intelligence and Data Mining.

Problem

Road damage in Simalungun Regency was logged manually and slowly, so hazards like potholes went unreported. The goal was an automated pipeline: capture → classify → locate → warn.

What I built

Results

The classifier reached strong accuracy on the held-out set and ran on-device via TensorFlow Lite. Full methodology and metrics are in the published article linked above.

AI/MLComputer VisionAndroidResearch