Let’s talk about how one of the world most luxurious and technologically advanced car companies, Tesla, is using machine learning to improve their autopilot system. Through the use of deep neural networks (DNN), Tesla is trying to classify and label the images that appear to the control when driving. It does this by creating an augmented training image for each image in a set of images. A neural network is the best application for this because it is an algorithm which functions to recognize and cluster input (data given to the autopilot system). In order to make the most effective technique it produces these augmented images for all eight different sensory cameras on the model. Each camera is positioned in a different location to maximize efficiency since more angles will allow for consistent results. Difficulties with training the computer model can be seen in the image below.
In this image to us, it may look like people walking in different directions towards or away from other objects. To a computer model it needs to account for the distances between the car and the subject, the environment (since its subject to change), and various other factors. One goal is to identify that the subject is a person and the store is a store and the car is for example, a police car with labels.To account for these problems, Tesla is still working on the best method but will experiment with, so far, three different embodiments. One, a non- transitory model that will provide instructions for the system through a processor. Two, having more than one non-transitory storage model along with a set of processors to compromise a set of instructions while in operation. And three, training the computer model through a set of parameters pre-determined towards the behavior of the autos environment. Although Tesla is still in the works of improving their autopilot feature, this company is revolutionizing both the field of the car-making industry and machine learning to provide a safer ride for their customers.