ParkSmart: Leveraging Neural Networks for Predictive Parking in Smart Cities

The surge in the proliferation of private vehicles within urban centers has exacerbated challenges in parking management, primarily due to a scarcity of available spaces. To address this issue, the implementation of intelligent parking systems featuring autonomous monitoring and guidance has become imperative. This paper proposes a manual model of ResNet-50 and other classification networks using the Global Perceptual Feature Extractor (GPFE) module. Evaluated on established datasets, namely PKLot and CNREXT, the GPFE module significantly enhances accuracy and resilience when integrated with CNN classifiers like ResNet-50. Additionally, the methodology of the paper employs cost-effective cameras for binary classification across diverse lighting scenarios. This innovative decision eliminates the need for Internet of Things (IoT) sensors, leading to substantial reductions in maintenance expenses. The integration of this module not only enhances the precision of binary classification but also addresses cost-related challenges, establishing it as a sustainable solution for parking problems.