YOLOv8 Object Detection Tutorial

Build a real-time object detection system using YOLOv8 and deploy it to edge devices with Cliff.

Overview

YOLOv8 is a state-of-the-art object detection model that can identify and locate objects in images and video streams. In this tutorial, we'll deploy YOLOv8 to an edge device for real-time inference.

Prerequisites

• Python 3.8+
• Basic knowledge of Python
• A Cliff account and registered device
• Camera or video source (optional)

Step 1: Prepare Your Model

Option A: Use Pre-trained Model

Download a pre-trained YOLOv8 model:

from ultralytics import YOLO

# Load a pre-trained model
model = YOLO('yolov8n.pt')  # nano version (fastest)
# or
model = YOLO('yolov8s.pt')  # small version
# or
model = YOLO('yolov8m.pt')  # medium version

Option B: Train Your Own

Train on custom data:

from ultralytics import YOLO

model = YOLO('yolov8n.pt')
results = model.train(data='path/to/dataset.yaml', epochs=100)

Step 2: Export to ONNX

Convert your model to ONNX format for deployment:

from ultralytics import YOLO

model = YOLO('yolov8n.pt')
model.export(format='onnx', simplify=True)

This creates a yolov8n.onnx file ready for deployment.

Step 3: Upload to Cliff

• Go to the Models page in your dashboard
• Click Upload Model
• Select your yolov8n.onnx file
• Set framework to "ONNX"
• Add metadata and upload

Step 4: Deploy to Edge Device

• Select your YOLOv8 model
• Click Deploy
• Choose your edge device
• Configure deployment:
• Set input size (e.g., 640x640)
• Configure confidence threshold
• Enable GPU if available
• Deploy

Step 5: Test Your Deployment

Once deployed, you can test with:

• Image files
• Video streams
• Camera feeds
• Real-time inference API

Performance Optimization

Model Quantization

Reduce model size and improve speed:

from ultralytics import YOLO

model = YOLO('yolov8n.pt')
model.export(format='onnx', simplify=True, int8=True)

Input Size

Smaller input sizes = faster inference:

• 320x320: Fastest, lower accuracy
• 640x640: Balanced (recommended)
• 1280x1280: Slower, higher accuracy

Use Cases

• Security monitoring
• Traffic analysis
• Quality control
• Inventory management
• Wildlife monitoring

Next Steps

• Experiment with different YOLOv8 variants
• Fine-tune on your specific use case
• Set up continuous monitoring
• Integrate with your application