This project showcases a lightweight and fast object detection pipeline using YOLOv8n — the smallest version of the Ultralytics YOLO family. It's ideal for real-time applications, experimentation, and learning purposes with small datasets like coco128.

• ✅ Based on YOLOv8n — "n" stands for Nano (smallest, fastest model)
• ⚡ Fast training and inference on low-resource machines
• 🗂️ Uses coco128, a small dataset subset for quick demo
• 📦 Built with Ultralytics YOLO
• 📈 Produces visual outputs (bounding boxes, metrics, logs)

YOLOv8n is the nano version of YOLOv8:
✅ Smallest model in the YOLOv8 family
✅ Designed for speed and low-latency inference
✅ Ideal for mobile, embedded, and edge devices

This example uses the coco128 dataset — a mini version of MS COCO for testing and demos.

Install dependencies:

pip install ultralytics

from ultralytics import YOLO
model = YOLO('yolov8n.pt')  # Nano version

results = model.train(data='coco128.yaml', epochs=5, imgsz=640)

results = model.predict(source='bus.jpg', conf=0.25)
for r in results:
    r.show()  # Display result

Sample prediction on bus.jpg using YOLOv8n:

After training and prediction, results and logs are stored in:

runs/detect/train/

Including:

• Training loss curves
• mAP charts
• Confusion matrix
• Detected image samples

• Want to use your own dataset? Replace coco128.yaml with your custom .yaml config.
• For even faster runs, lower epochs or reduce image resolution with imgsz.

This project uses the Ultralytics YOLO License.