#55DAC 2: Teams tackle deep learning in DAC design contest
December 13, 2017 by Sharon Hu - General Chair DAC55 and a professor in dept of CS and Engineering at Univ of Nortre Dame, Indiana
When it comes to tackling leading-edge design challenges in fun ways, there’s no better place than DAC. For DAC 2018, we’ve created a System Design Contest targeting machine learning on embedded hardware.
If you think this is too leading edge for a design contest, you’d be mistaken: More than 100 teams registered for the contest. You can find a full list of the teams here: http://www.cse.cuhk.edu.hk/~byu/2018-DAC-HDC/teams.html
So how does the contest work:
Teams had the choice of using a Xilinx PynQ-Z1 FPGA-based development system or an NVIDIA Jetson TX2 development system, as well as software and deep-learning tools kits. Xilinx and NVIDIA donated boards to support the efforts.
Drone maker DJI donated a data-set that included more than 100 video clips with full annotation of the bounding box for the tracking object (a person or car).
The teams built either FPGA- or GPU-based systems to track people and vehicles from consumer drones using deep learning methods running on advanced embedded systems platforms. A hidden dataset is used to evaluate the performance of the designs in terms of accuracy and power consumption.
I can’t over-state the energy and generosity of the contest sponsors—NVIDIA, Xilinx, DJI, ACM/SIGDA and DAC— in this endeavor. Thank you! It’s so rewarding to see industry and innovators come together to enable superb system design.
It’s been an honor to work with industry luminary Chris Rowen, and the University organizers Yiyu Shi at Notre Dame, Jingtong Hu at University of Pittsburgh and Bei Yu at the Chinese University of Hong Kong on this project.
The teams will be posting results at the end of each month, starting in late January, so you can follow their progress leading up to DAC. Mark your calendars and follow the teams as they design to win cash awards sponsored by Nvidia and Xilinx. Awards will be given to the top three teams in each category, and the top winners will be on hand in San Francisco at DAC to present their work.