metadata
license: mit
metrics:
- precision
- recall
base_model:
- Ultralytics/YOLOv8
pipeline_tag: object-detection
tags:
- defect-detection
- industrial, quality-control, yolov8
Industrial Surface Defect Detection Model (NEU-DET)
A YOLOv8-based deep learning model for real-time detection of 6 types of surface defects in industrial materials.
📋 Model Overview
Architecture: YOLOv8
Dataset: NEU-DET (1,800 grayscale images)
Task: Object Detection (Defect Classification)
Framework: Ultralytics
Input: Images (JPEG/PNG)
Output: Bounding boxes + Confidence scores
🔍 Supported Defect Classes
| Defect Type | Description |
|---|---|
| Crazing | Fine surface cracks forming network patterns |
| Inclusion | Foreign material embedded in surface |
| Patches | Surface irregularities and discoloration |
| Pitted Surface | Pitting and corrosion damage |
| Rolled-in Scale | Scale/oxide layers rolled into material |
| Scratches | Surface abrasions and scratch marks |