Comprehensive Study of YOLO Versions for Front and Rear-View Classification of Vehicles in Context of Indian Roads.

Ever since Computer Vision was introduced, humanity has seen various ways to detect or classify objects of various types. Depending upon the context in consideration, the performances of models vary with respect to their evolution or even upon the nature of the data in hand. The classification of front or rear views in vehicles forms an integral part when we go ahead with deciding whether a given vehicle is moving in the correct lane. In the context of Indian streets, we have various challenges like rural unmarked roads, faded markings, shaded situations from poles or trees, etc. Hence instead of detecting lanes, an alternative way is to detect whether the vehicle(s) ahead is facing toward or away from our vehicle. Various deep learning architectures have been proposed in this aspect to detect or classify objects like the networks from Visual Geometry Group, You Only Look Once, Inception Networks, Residual Networks, etc. In this paper, we have performed a comparative analysis of performance on various versions of You Only Look Once for its evolution over time. [ABSTRACT FROM AUTHOR]