Vision transformers started making waves in
deep learning by replacing the typical convolutional neural
networks (CNN) in tasks like image classification, Object
detection, segmentation etc. the implementation of vision
transformers can be further extended to autonomous cars.
As it was proven that pure transformer architecture can
outperform CNNs when trained over a large dataset by
using comparatively less computational resources .
These vision transformers can be implemented in selfdriving cars to calculate the optimal steering angle by
capturing images of the surroundings. The vision
transformers can take advantage of the attention
mechanism to focus on the most important things in the
image like road lanes, other vehicles, traffic signs, etc. and
produce better results compared to CNN models. The
paper discusses about implementation behavior cloning
using vision transformers in a self-driving car which is
simulated in a Udacity self-drive car simulator.
Keywords : Transformers, Self-Driving Vehicles, Vision Transformers, Convolutional Neural Networks CNN, Behavior Cloning.