Parking space classification is crucial in reducing traffic congestion, curbing air pollution, and enhancing driver convenience. This research presents a robust model combining a Convolutional Neural Network (VCNN) variant, specifically ResNet50, with the Customized Vision Transformer (VIT) for efficient feature extraction and classification. The model performs better by integrating ResNet50’s exceptional feature extraction capabilities with VIT’s global semantic information. VIT is customized to adapt to the extracted feature vectors. The model’s efficacy is rigorously evaluated using the widely recognized PKLot dataset, demonstrating its proficiency in distinguishing between empty and occupied parking spaces through precision, sensitivity, specificity, accuracy, and F1-score metrics. Furthermore, benchmarking against existing deep learning models highlights its superiority. This innovation promises significant advancements in intelligent transportation systems, providing an efficient parking space classification solution that mitigates traffic congestion, prevents unauthorized parking, and fosters brighter urban environments.