road-damage-detection

Detect road anomalies such as cracks, potholes, and bumps using our trained YOLOv8 models with visual demo. Real-time detection via Streamlit and Flask app

Road Damage Detection Based on Unsupervised Disparity Map Segmentation (T-ITS)

Integrated real-time data analytics for optimized public transport, innovative road monitoring using demand prediction, and conditioning tech for sustainability, real time pothole detection either by image or video, smart parking count system for efficiency using AI/ML.

Detect damages like cracks and potholes of roads

基于YOLO模型的路面缺陷检测（Road Damage Detection）

YOLOv7-based road damage detection system trained on RDD2022 multi-national dataset. Achieves 63.85% mAP for detecting cracks and potholes. Includes Gradio UI.

An application for automatic road damage assessment using semantic segmentation on high-resolution images. The project helps municipal authorities and maintenance teams detect and prioritize road repairs, improving safety and reducing costs.

Unity-based digital twin for autonomous UAV pavement inspection under dynamic traffic, with occlusion recovery and multitask YOLOv8 perception.

YOLO-based road anomaly detection project with a Streamlit app, packaged model weights, archived evaluation artifacts, and ready-to-run local setup.

PyTorch implementation of EfficientDet (D0, D2, D3) for Road Damage Detection on RDD2022 dataset. Includes K-Medoids anchors, DDP training, and COCO evaluation.

Real-time pothole detection Android application using computer vision, OpenCV, and native C++ processing. Developed during a designathon that led to an Ericsson internship.

AI-powered ADAS pothole detection system using YOLO, Random Forest, Flask and OpenCV.

Baseline implementation of Road Damage Object Detection using YOLOv8, YOLOv10, and YOLO11 with Hyperparameter Tuning and Streamlit deployment.

An road-anomaly detection system designed to automatically identify various types of road damage

This project is an AI-powered pothole monitoring platform that detects road defects using YOLO, allows citizens to report issues, and provides administrators with a dashboard to track and resolve them. It combines computer vision, backend automation, and user interfaces to make road maintenance faster and smarter.

Transfer learning benchmark: ResNet-50, EfficientNet-B3, ViT-B/16, MobileNet-V3 on RDD2022 Japan pavement distress dataset — accuracy, calibration, latency, and parameter count evaluated.

Deep Learning-based road damage detection and classification. Features a CNN pipeline for image analysis, a web-based monitoring dashboard, and an integrated messaging service for real-time safety alerts.

Reproducible CPU-only road damage detection on RDD2022 with domain-aware COCO evaluation and an IEEE paper

AI-powered road damage detection and road health assessment using YOLO, Streamlit, and GPS visualization.

Real-time road damage detection on Raspberry Pi 5 using YOLOv8 Nano with NCNN and Hailo-8 backends. Engineering thesis — AGH UST 2026.