Overview:
The Crack Segmentation project is aimed at detecting road potholes using computer vision to improve road conditions. The system is built using YOLOv8l, and it processes a dataset of 1551 images to identify cracks, contributing to road maintenance and safety efforts.
Key Features:
1. YOLOv8l Model for Crack Detection:
This project utilizes YOLOv8l for precise detection of road cracks, optimizing the model for crack segmentation within the dataset.
2. Dataset:
The model was trained on 1551 images with a resolution of 640×640, specifically targeting one class: cracks, to accurately identify potholes.
3. Training Process:
The data was split into 70% for training, 20% for validation, and 10% for testing, with 40 epochs used to fine-tune the model.
4. Performance Metrics:
Achieved a mean average precision (mAP) of 0.824, with precision at 0.8 and recall at 0.829, showcasing its reliability in detecting cracks.
5. No Data Augmentation: T
he model was trained without any augmentation, indicating strong baseline performance even without additional preprocessing techniques.
Category:
Tags:
Links
Portfolio 
Leave a comment