YOLOv11-Nano Car Damage Detection

A lightweight YOLOv11-Nano model fine-tuned for automated detection and classification of automobile damage across 14 damage categories.

⚙️ Model Details

• Architecture: YOLOv11-Nano
• Framework: Ultralytics
• Input Size: 640x640
• Format: PyTorch (.pt)

📊 Model Performance

• mAP@0.5: 79.7%
• mAP@0.5:0.95: 55.2%
• Model Size: ~6MB (Nano variant)
• Speed: Real-time inference capable

🎯 Use Cases

• Insurance claim automation
• Vehicle inspection systems
• Auto repair assessment
• Fleet damage monitoring

📁 Dataset

Trained on the Automobile Damage Detection Dataset from Roboflow Universe.

• Images: ~6,900 annotated images
• Classes: 14 damage categories
• Annotations: Bounding boxes

Damage Categories

• Front-windscreen-damage
• Headlight-damage
• Rear-windscreen-Damage
• Runningboard-Damage
• Sidemirror-Damage
• Taillight-Damage
• bonnet-dent
• boot-dent
• doorouter-dent
• fender-dent
• front-bumper-dent
• quaterpanel-dent
• rear-bumper-dent
• roof-dent

📈 Training Metrics

Results Plot

Confusion Matrix

🚀 Usage

from ultralytics import YOLO
from huggingface_hub import hf_hub_download

model_path = hf_hub_download(
    repo_id="vineetsarpal/yolov11n-car-damage", 
    filename="best.pt"
)
model = YOLO(model_path)

results = model.predict("/path/to/your/image.jpg", save=True)

⚠️ Limitations

• Class imbalance: Reduced accuracy on boot-dent detection due to limited training examples
• Image quality: Performance degrades on low-resolution (<640px), poorly-lit, or low-contrast images