A Parallel Weight Aggregated Ensemble of pre-trained models for Open Parking Space Detection

Emerging data on urbanization projects highlights a significant trend: By the foreseeable future, an estimated 70% of the global population is projected to call urban areas home. [1]. Hence smart city initiatives have become imperative for a smart living in cities with smart mobility extremely critical. Smart parking is an essential component in smart mobility which assists the vehicle drivers in making parking decisions smartly and efficiently. These systems aim to provide real time free parking space information accurately and route them to the slot in lesser time and distance. Deep learning methods have demonstrated their effectiveness in the majority of vision-based solutions for open parking space detection. The proposed work compares cutting-edge deep learning techniques used for detection of empty and occupied parking lots in open parking areas. It offers its readers a relative evaluation of the current deep learning models. It evaluates and concludes that the accuracy of pre-trained ensemble based learning models outperforms performance and robustness of the established solutions and work well on new and unseen data.