Research on verificiation of deep learning in autonomous cars.
One piece of the puzzle in making sure self-driving vehicles are safe to use is the verification of deep learning in safety critical applications. This is the focus of Semcon’s industrial doctoral student specialized in AI, Jens Henriksson.
In the past decade, automated vehicles have transitioned from an industry vision to fleet tests with small or limited deployment. This transition has been possible thanks to advances in the field of computer vision, aided by advances with deep learning.
Incorporating deep learning into safety critical applications comes with inherent challenges that need to be addressed before large scale deployment can be achieved. This thesis has investigated what additional measures of testing are needed for deep neural networks, and studied the challenge of detecting out-of-distribution samples more in depth.
Read the full thesis here: “On Improving Validity of Deep Neural Networks in Safety Critical Applications”.