Regularização em Redes Neurais com Reamostragem de Dados para Monitoramento Urbano Inteligente

The advancement of smart cities has led to a growing demand forefficient urban management systems, such as parking space management.In this context, regularization has proven to be an essentialtechnique for improving the generalization of deep learning models.This paper proposes a regularization method based on the use of smallerdata subsets across multiple epochs with resampling, aiming tobalance model learning and reduce the risk of overfitting. Using theMobileNetV3-Small, four experiments were conducted with the PKLotand CNRPark-EXT datasets. The results indicate that the proposedmethod achieved competitive results while utilizing only a fraction ofthe original data during training. By employing these strategies, itwas possible to reduce the data required per epoch by up to 97%, whilemaintaining an average accuracy close to 89%. Furthermore, trainingwith the PKLot dataset highlighted the positive impact on model robustnesswhen using datasets with greater diversity and quantity. Thisstudy underscores the importance of new regularization approachesto enhance the efficiency and generalization of deep learning modelsin urban applications.