Review of Fault Mitigation Approaches for Deep Neural Networks for Computer Vision in Autonomous Driving

The aim of this work is to identify and present challenges and risks related to the employment of DNNs in Computer Vision for Autonomous Driving. Nowadays one of the major technological challenges is to choose the right technology among the abundance that is available on the market. Specifically, in this thesis it is collected a synopsis of the state-of-the-art architectures, techniques and methodologies adopted for building fault-tolerant hardware and ensuring robustness in DNNs-based Computer Vision applications for Autonomous Driving.