Autonomous vehicles are already a reality, and there are still severalchallenges to overcome. One important challenge for the adoptionof these vehicles is perceiving its surroundings. This necessity ofperception can be fulfilled by digital cameras. When working withdigital image processing, the quality will be limited by real-timeconstraints. As several works indicate, this real-time constraint forautonomous vehicles is at most 100ms per frame. Also, by improvingthe processing time, the chances of accidents involving autonomousvehicles may be decreased. This paper analyses the advantages anddrawbacks of semantic segmentation and also presents a study toimplement perception for autonomous vehicles by accelerating asemantic segmentation algorithm, also used by other works on thefield. To accelerate the algorithm, spacial parallelism will be used.
Paper
Full text
A Study on Semantic Segmentation for Autonomous Vehicles
OpenAlex · Advanced Neural Network Applications · 2020
Abstract
Autonomous vehicles are already a reality, and there are still severalchallenges to overcome. One important challenge for the adoptionof these vehicles is perceiving its surroundings. This necessity ofperception can be fulfilled by digital cameras. When working withdigital image processing, the quality will be limited by real-timeconstraints. As several works indicate, this real-time constraint forautonomous vehicles is at most 100ms per frame. Also, by improvingthe processing time, the chances of accidents involving autonomousvehicles may be decreased. This paper analyses the advantages anddrawbacks of semantic segmentation and also presents a study toimplement perception for autonomous vehicles by accelerating asemantic segmentation algorithm, also used by other works on thefield. To accelerate the algorithm, spacial parallelism will be used.