Logo Lanfrica

Tosinvincible2017/vnpr-pavement

Type de record:

software
Créateur:
Tos
Hôte:
YOLOv8 and EasyOCR-based real-time vehicle and license plate recognition system with Supabase database management. Traffic data (ADT & ESAL) collected from Nsukka–Enugu Road is used to inform road pavement thickness design. Built with Python, Google Colab, and a mobile app dashboard. # VNPR-Pavement: Integrated Vehicle & Number Plate Recognition for Road Pavement Design > A deep learning system using **YOLOv8** and **EasyOCR** for real-time vehicle and license plate detection, backed by **Supabase** cloud database management — with traffic data (ADT & ESAL) used directly to inform **road pavement thickness design**. Developed at the University of Nigeria, Nsukka. --- ## System Overview ``` Video Feed (Camera) │ ▼ [YOLOv8 Detection] Vehicle + Plate │ ▼ [EasyOCR — OCR] Alphanumeric Text │ ├──────────────────────────┐ ▼ ▼ [Supabase Database] [Traffic Analysis] License plates ADT · ESAL · Load Timestamps Classification Vehicle types │ ▼ [Pavement Thickness Design] AASHTO · Nsukka–Enugu Road │ ▼ [Mobile App Dashboard] ``` --- ## Key Technologies | Component | Technology | |---|---| | Object Detection | YOLOv8 (Ultralytics) | | OCR | EasyOCR | | Database | Supabase (PostgreSQL) | | Training Platform | Google Colab (NVIDIA T4 GPU) | | Language | Python | | Mobile App | Flutter / React Native | --- ## Dataset - **Kaggle** large license plate dataset — diverse lighting, angles, and conditions - **Real-time** footage from Nsukka–Enugu Road, Nigeria - Training: 8,922 images | Validation: 1,912 images | Test: 1,912 images (70/15/15 split) --- ## Results — Test Images ### Plate Detection ### Character-Level Accuracy Distribution ### Common Character Errors --- ## Results — Real-Time Images (Nsukka–Enugu Road) ### Plate Detection ### Character-Level Accuracy Distribution ### Worst Performing Samples --- ## Mobile Application --- ## Ground Truth Files | Test Images | Real-Time Images | |:---:|:---:| | | | --- ## Pavement Design Integration Traffic data captured by the VNPR system feeds directly into pavement engineering calculations: ### Average Daily Traffic (ADT) Continuous vehicle counts from the VNPR system replace manual counting, providing more …