Markus Tremmel, Chief Expert for ADAS at Bosch, presents the “Computer Vision Hardware Acceleration for Driver Assistance” tutorial at the May 2018 Embedded Vision Summit.
With highly automated and fully automated driver assistance system just around the corner, next generation ADAS sensors and central ECUs will have much higher safety and functional requirements to cope with. This trend directly translates into a huge uptake of the required calculation performance and hence into much higher power consumption and system cost. Due to the increased amount and complexity of visual sensors around the car, the embedded computer vision subsystem carries a major stake of this increase.
To realize efficient, safe and affordable L3 and higher ADAS solutions, it is important to use the best possible compromise of hardware acceleration between fixed logic and specialized processing architectures. This presentation gives an overview of the different computer vision method options available in a next-generation ADAS system and looks at the trade-offs between fixed logic and programmable processing units with an eye on the latest developments in deep learning.