Efficient Parking Space Detection with Multi-Scale Feature Networks and YOLOv5

Efficient parking management is a critical challenge in modern urban environments, driven by increasing vehicle populations and limited parking spaces. This research proposes a novel deep learning-based framework for real-time parking space detection, leveraging the YOLOv5 architecture enhanced with multi-scale feature pyramid networks and attention modules. The framework aims to accurately detect parked vehicles and empty parking spaces under diverse conditions, such as varying lighting, vehicle types, and parking orientations. To address the scarcity of suitable datasets, we developed a comprehensive parking lot image dataset captured using CCTV cameras, ensuring robust performance in real-world scenarios. Experimental results demonstrate the proposed model's effectiveness, achieving high accuracy and real-time performance. Comparative analysis highlights the model's superior capabilities in detecting compact vehicles and managing challenging lighting conditions compared to existing approaches.