DefectVision — YOLOv8 Defect Detection (bottle)
YOLOv8n fine-tuned on MVTec AD dataset for industrial defect detection.
Category : bottle
- Defect types : broken_large, broken_small, contamination
- Classes : 1 (defect)
- Image size : 640x640
Metrics (test set)
- mAP50 : 0.0933
- mAP50-95 : 0.0305
- Precision : 0.0028
- Recall : 0.8095
Usage
from ultralytics import YOLO
from huggingface_hub import hf_hub_download
path = hf_hub_download(repo_id='Chasston/defect-vision-yolov8-bottle', filename='best.pt')
model = YOLO(path)
results = model.predict('your_image.png', conf=0.25)