Learning Approaches for Parking Lots Classification

Daniele Di Mauro, Sebastiano Battiato, Giuseppe Patanè, Marco Leotta, Daniele Maio, Giovanni Maria Farinella

Abstract: The paper exploits the problem of empty vs. non-empty parking lots classification from images acquired by public cameras through the comparison between a classic supervised learning method and a semi-supervised learning one. Both approaches are based on convolutional neural networks paradigm. Experimental results point out that the supervised method outperforms the semi-supervised approach already when few samples are used for training.