Smart Parking Management System

Real-time parking occupancy detection using YOLOv8, Kafka, MongoDB, and Flask

---

Overview

A real-time smart parking platform that combines computer vision and event-driven data engineering to monitor parking space availability. The system uses a YOLOv8-based object detection pipeline to analyze camera images, streams occupancy events through Apache Kafka, stores them in MongoDB, and serves live availability data via a Flask web dashboard that auto-refreshes every 5 seconds.

---

Architecture

Camera Images
     │
     ▼
[YOLOv8 Model] ──── occupancy detection
     │
     ▼
[Kafka Producer] ──── streams events to topics
     │
     ├──── IoT sensor topic ──► [Consumer] ──► MongoDB (parking_data)
     │
     └──── Image topic      ──► [Consumer] ──► MongoDB (image_stats)
                                                    │
                                                    ▼
                                            [Flask Dashboard]
                                         (auto-refresh every 5s)

---

Features

• YOLOv8 computer vision — detects occupied/free parking spaces from images using a fine-tuned YOLO model (best.pt)
• Apache Kafka streaming — event-driven architecture with separate producer/consumer pairs for IoT sensor data and image analysis
• MongoDB storage — persists occupancy events with timestamps for both sensor and vision pipelines
• Flask web dashboard — displays real-time available spaces from both data sources, with automatic 5-second refresh
• Docker Compose — full containerized deployment of Kafka and MongoDB

---

Tech Stack

Component	Technology
Object Detection	YOLOv8 (Ultralytics), ONNX
Message Streaming	Apache Kafka
Database	MongoDB
Web Framework	Flask
Containerization	Docker Compose
Language	Python 3.7+

---

Project Structure

car_parking/
├── app.py                # Flask web server and dashboard
├── producer.py           # Kafka producer for IoT sensor data
├── producer_images.py    # Kafka producer for camera images
├── consumer.py           # Kafka consumer → MongoDB (parking_data)
├── consumer_image.py     # Kafka consumer → MongoDB (image_stats)
├── image_to_data.py      # Image preprocessing pipeline
├── kafka_to_mongo.py     # Kafka–MongoDB bridge utilities
├── ml_model.py           # YOLOv8 inference wrapper
├── test.py               # Testing utilities
├── best.pt               # YOLOv8 fine-tuned model weights
├── best.onnx             # ONNX export of the model
├── docker-compose.yml    # Kafka + MongoDB containers
├── Templates/            # Flask HTML templates
└── images/               # Sample parking lot images

---

How to Run

Prerequisites

• Python 3.7+
• Docker and Docker Compose

1. Clone the repository

git clone https://github.com/karanehzm/car_parking.git
cd car_parking

2. Install Python dependencies

pip install flask pymongo kafka-python ultralytics opencv-python

3. Start Kafka and MongoDB containers

docker-compose up --build -d

4. Run the pipeline (in separate terminals)

# Start IoT data producer
python3 producer.py

# Start image producer
python3 producer_images.py

# Start IoT data consumer
python3 consumer.py

# Start image consumer
python3 consumer_image.py

# Start Flask web server
python3 app.py

5. Open the dashboard

Navigate to http://localhost:5000

---

Dashboard Preview

---

Author

Karaneh Zolfaghari Moghaddam
github.com/karanehzm