“…The main difference between state-of-the-art OSPI models available and how they are validated is the data gathered and the features considered for training, validating, and testing the models [19]. Data sources that have been used to validate parking prediction models are: smart parking meters [15,18,20,21], mobile payments [8,22,23], intelligent parking systems [24], real-time ground sensors [14,17,25,26] images captured by a camera mounted on a moving vehicle [7,27], crowd-sensing information by equipping probe vehicles (e.g. taxis) with onboard sensors, cameras, or ultrasonic sensors [28,29], or crowdsensing using GPS signals from smartphones [23,[29][30][31], and also manual observations [32].…”