Deep Learning for Automotive - A Heterogeneous Approach
The Artificial Intelligence is revolutionizing our world. Exponential growth in processing power in silicon and simultaneous reduction in its cost has created a new class of embedded technologies that have built-in Convolutional Neural Networks (CNNs). The CNN architecture is inspired by the neurons of the human brain. This is also known as ‘Deep Learning’, essentially because, layers of ‘Deep Neural Networks’ are trained using data sets from the real world. The ultimate goal here, is to empower the CNN to receive real time inputs from the environment and enable it to provide ‘outputs’ without any ‘errors’. This process is iterative and intense. CNN consumes Petabytes (and more) worth of data in order to achieve acceptable levels of predictive decisions.
Deep Learning can be applied to ‘image’, ‘voice’ and ‘sensory’ datasets to enable seeing and sensing autonomous vehicles. HCLTech’s white paper discusses the various approaches of using CNN to create solutions for Automotive use cases in image recognition. We hope you find it useful.