Although not much has been spoken about Apple’s self-driving project by analysts, we at least know that it exists and is being worked upon. Rumours from past years indicated the tech titan to be working on a full-blown automobile, but in recent times this piece of information has taken a different direction. From last year it was reported that Apple has shifted its focus from developing an automobile to developing self-driving software.
This was almost confirmed when Apple reportedly received the permit from California DMV to test self-driving cars in April this year. The rumours were further cemented when the man himself, CEO of Apple, Tim Cook himself revealed in a talk show in June that the company is indeed working on software for self-driving cars.
And just before the year comes to an end, several new details have been confirmed by the company’s AI research director, Ruslan Salakhutdinov.
Salakhutdinov, while speaking to approximately 200 AI experts during the Neural Information Processing Systems (NIPS) conference in Long Beach, California, talked about Apple’s advances in the field of AI and machine-learning, most of which was in relation to self-driving cars.
He talked about how Apple is using data gathered by cameras and other sensors and made 3D maps of cities. Salakhutdinov showed a project that was revealed in a project published by Cornell University. The project detailed the technology that trains software in identifying pedestrians and cyclists with LiDAR, a 3D scanner, which is used in several autonomous cars.
He also talked about other projects that don’t seem to have been published anywhere else, as mentioned by Wired. The Apple executive talked about a software that identifies cars, pedestrians, road lanes and more using a camera or more than one cameras in an automobile.
Salakhutdinov even showed images suggesting how the system was able to recognise pedestrians, sidewalk and more even when the raindrops covered the camera lens. Another project talked about was regarding how a software uses a technique called SLAM to get a sense of direction. The technique is generally used in robots and autonomous vehicles for object mapping purposes. One of the projects was in relation to sensor-laden cars to develop a 3D map along with features such as traffic lights and more such elements.
He also mentioned about how a software can make decisions by itself in ‘dynamic’ situations, a diagram in one of his slides reportedly showed a car plotting its way around a pedestrian.