Development Environment Proposal for Automotive Vision-Based Algorithms

Next generation autonomous driving functions heavily depend on various computationally intensive algorithms based on machine vision. Design and development of such algorithms is challenging, requiring engagement of a wider community and the availability of appropriate, cost-effective tooling. Setting up a hardware and software environment that would both reflect actual automotive requirements and facilitate development and prototyping, is an uneasy task. In this paper, we analyze the available development platforms which could be utilized in automotive machine vision, with specific regard to new algorithm requirements, such as the utilization of AI, hardware acceleration and low latency. We give a proposal of an affordable and scalable development setup consisting of heterogeneous, vision-enabled SoCs and automotive interfaces.