Explainable Real Time Vehicle Make Model Methodology Using Bag of Expressions Features for Indian Passenger Cars

Passenger Vehicle Identification is useful in many of the general-purpose applications. It will improve Intelligent transport system. It will also help to implement law and surveillance.  There are many of the deep learning approaches to identify vehicles.  Major work done based on shape, and back light identification. Another issue is research work is done on Europe and American vehicle’s so there is lot of scope of work for the Indian vehicles.

Vehicle identity is also degraded with speed, day night conditions, rain etc. In addition to it the road traffic conditions are based on locations. It carries with weather and location. This research work is based on identification of the many features of the Vehicle and then comparing and take a decision which is based on multiple features. This approach decomposes area of the interest in the vehicles into small area of expressions. This includes identification of the headlamps, grille, logos, fog lamps, bumpers, and back lights. YOLOv8 framework and EfficientNet-B4 is used to extracts features, then train them and identify the vehicle. In this work a new data set for the Indian passenger vehicle have been planned and work is in progress to create a new data set for the work. This experimental result is under progress at this stage