Parking Spaces Detection Using Machine Learning

Project Overview

Our recent project focused on developing a Parking Spaces Detection system using machine learning. The goal was to create an efficient and cost-effective solution to monitor parking spaces in real-time, ensuring optimal utilization and management.

Technical Details

• Hardware: Initially, we considered using high-end GPUs for real-time video analysis. However, after understanding the client’s requirement for snapshot analysis every 5 minutes, we opted for a more cost-effective solution using a Raspberry Pi with a higher configuration.
• Software: The core of our system is built using Detectron2, a powerful library for object detection. We utilized OpenCV for image processing and PyTorch for model implementation.
• Model: The model was trained to detect parking spaces from static images. It takes a snapshot every 5 minutes and processes it to identify available and occupied parking spaces.
• Deployment: The system is designed to run on an OAK-D device, ensuring efficient processing and easy integration with existing infrastructure.

Achievements

• Cost-Effective Solution: By switching from a high-end GPU to a Raspberry Pi, we significantly reduced the project’s cost without compromising performance.
• Efficient Detection: The model accurately detects parking spaces from static images, providing reliable data for parking management.
• Scalability: The system can be easily scaled to monitor multiple parking lots, making it a versatile solution for various applications.