Abstract
Smart parking systems having efficient parking vacant spot detection and vehicle entrance and exit count can be quite beneficial for managing traffic and it can play a vital role for reducing cost of fuel. Parking spot detection using visual streams from cameras can be quite efficient as it does not require any sensor to be installed on parking location separately. In this paper, we proposed a unique parking vacancy slot detection using state-of-the-art detection of vehicles based on Faster R-CNN. Furthermore, vehicle entrance and exit count is also determined based on deep convolution features. The proposed system is evaluated on publically available PKLot dataset and achieved improved accuracy of 8% as compared to baseline methodology.