Efficient Smart Parking Space Identification and Classification System Using YOLOv8 Network Model

In urban areas, the increasing number of vehicles has resulted in significant challenges related to parking management, such as inefficient space utilization, increased traffic congestion, and difficulty locating available parking spots. Traditional parking systems are often inefficient and fail to provide real-time, accurate data about parking space availability. The project aims to develop a car parking availability detection system using the Convolution Neural Networks model. The system will provide real-time, highly accurate parking space detection by leveraging computer vision and advanced object detection techniques. The system provides the count of both occupied and empty parking spaces, displayed on the output image or video feed, making it particularly beneficial for integration into smart city infrastructure to optimize parking resource utilization, reduce congestion, and improve user convenience. The proposed system, a custom dataset consisting of 13,786 images, was created, capturing various parking lot scenarios with different occupancy conditions. Each image is labeled to denote parking spaces and their respective occupancy statuses. Three subsets of the dataset have been created: 10% for testing, 20% for validation, and 70% for training. Data augmentation methods like image flipping, rotation, and scaling were used during training to improve the model's resilience in various environmental circumstances. The model's efficacy is asses using numerous metrics, including accuracy, F1 score, and the confusion matrix. The proposed Customized YOLOv8 Model demonstrated high accuracy, achieving training accuracy of 99% and validation accuracy of 97%.