👁️ YOLOv8 Real-time MLOps Service

A production-ready Computer Vision service designed for real-time object detection. It features a scalable FastAPI backend, Dockerized deployment, CPU optimization algorithms, and a full observability stack (Prometheus & SQLite logs).

Note: This dashboard demonstrates real-time inference capabilities with minimal latency.

🚀 Key Features

• ⚡ Real-time Inference: Streams processed video via MJPEG with YOLOv8 bounding boxes.
• 🛠️ MLOps Architecture:
  • Fully Dockerized for easy deployment.
  • Configuration Management via .env files (12-factor app).
  • CI/CD Ready structure.
• 📉 CPU Optimization: Implemented "Smart Throttling" algorithm to maintain stable 30 FPS and prevent CPU overheating on standard hardware.
• 📊 Observability:
  • Prometheus Metrics (/metrics) for monitoring request latency and throughput.
  • SQLite Logging for tracking detection history on uploaded images.
• 🛡️ Security: Stream endpoints hidden from Swagger documentation to prevent service overload.
• 🎨 Interactive Dashboard: Custom Dark Mode UI for monitoring system status and video feed.

🛠️ Tech Stack

Component	Technology	Description
Framework	FastAPI	High-performance async web framework.
Model	YOLOv8 (Ultralytics)	State-of-the-art object detection model (yolov8n.pt).
Vision	OpenCV	Image processing and video stream handling.
Database	SQLite	Lightweight storage for detection logs.
Container	Docker	Multi-stage build based on python:3.12-slim.
Metrics	Prometheus	Instrumentator for exposing app metrics.

🏗️ Project Structure

.
├── data/                  # Persisted data (SQLite DB)
├── main.py                # Application entry point & logic
├── Dockerfile             # Docker image configuration
├── requirements.txt       # Python dependencies
├── .env.example           # Template for environment variables
└── README.md              # Project documentation

⚡ Quick Start

Prerequisites

• Docker & Docker Compose installed.
• Git.

1. Clone the Repository

git clone https://github.com/Western-1/YOUR_REPO_NAME.git
cd YOUR_REPO_NAME

2. Configure Environment

cp .env.example .env
# You can modify VIDEO_SOURCE in .env to use a YouTube URL, RTSP stream, or local file.

3. Build & Run with Docker

# Build the image
docker build -t yolo-service .

# Run the container (Map port 8000 and mount data volume)
docker run -d -p 8000:8000 -v $(pwd)/data:/app/data --env-file .env --name yolo-app yolo-service

4. Access the Service

• Dashboard: http://localhost:8000
• API Documentation: http://localhost:8000/docs
• Metrics: http://localhost:8000/metrics

📡 API Endpoints

Method	Endpoint	Description
GET	/	Main Dashboard (HTML).
POST	/detect_image	Upload an image, get it back with bounding boxes + Log to DB.
POST	/detect_json	Upload an image, get JSON stats (e.g., {"person": 3, "car": 1}).
GET	/history	View recent detection logs from the database.
GET	/metrics	Prometheus scraping endpoint.
GET	/video_feed	(Internal) MJPEG video stream.

⚙️ Configuration (.env)

Variable	Default	Description
APP_NAME	YOLOv8 MLOps Service	Name displayed on Dashboard.
MODEL_PATH	yolov8n.pt	Path to the YOLO weight file.
VIDEO_SOURCE	(Intel Sample Video)	URL to video file, YouTube link, or RTSP stream.
MAX_LOG_ENTRIES	1000	Database rotation limit.

---

👨‍💻 Author

Developed by Andriy Vlonha (@Western-1).

📝 License

MIT License - see LICENSE file for details.

Copyright (c) 2026 Andriy Vlonha

---

📞 Contact

Andriy Vlonha 📧 Email: [email protected] 💼 LinkedIn: Andriy Vlonha 🐙 GitHub: @Western-1 📱 Telegram: @Westerny

---