metadata
license: apache-2.0
tags:
- yolov8
- image-classification
- medical
- wound-detection
- smart-triage
datasets:
- yasinpratomo/wound-dataset
- ibrahimfateen/wound-classification
- shubhambaid/skin-burn-dataset
metrics:
- accuracy
pipeline_tag: image-classification
Smart Triage Ambulance — Wound Classification Model
YOLOv8n-cls fine-tuned on multi-source wound imagery for real-time surface injury classification in ambulance triage scenarios.
Classes
| Class | Description |
|---|---|
wound |
Open injuries — lacerations, cuts, abrasions, stab wounds |
burn |
Thermal damage — 1st, 2nd, 3rd degree burns |
contusion |
Closed injuries — bruises, swelling without skin break |
Training Details
- Base model: YOLOv8n-cls (ImageNet pretrained)
- Dataset: 3 Kaggle sources, quality-audited (blur, resolution, dedup), class-balanced via ambulance-specific augmentation
- Optimizer: AdamW, lr=0.001, freeze=10 layers, dropout=0.3
- Image size: 224×224
Usage
from ultralytics import YOLO
model = YOLO("best.pt")
results = model.predict("wound_image.jpg")
print(results[0].names[results[0].probs.top1])
Intended Use
Prototype component for the Smart Triage Ambulance system. Not for clinical diagnosis. Designed for real-time paramedic decision support during patient transit.