Abstract
We report on experiences with optimizations that enable neural network edge computing on a small robot vehicle, which serves as a proxy for an autonomous robotic space craft. We realize a visual object recognition task using a neural network on a field-programmable gate array (FPGA) with data processing resources as limited as those of an FPGA suitable for space. We use a quantized neural network nicely matching the properties of an FPGA. The restrictions of the small FPGA require us to sequentialize the processing partially. We employ input frame tiling for this. It allows us to keep the entire neural network on-chip. Furthermore, we split up the visual object recognition task into two stages, using two separate neural networks. The first stage identifies the region of interest approximately, using large and thus few tiles. The second stage looks closely at the single tile containing the region of interest; thus being not that time critical.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.