This special session highlights the winning entries of the 2019 DAC System Design Contest on Low Power Object Detection (SDC). The contest started late last year. Contestants were required to implement object detection machine learning algorithms in either FPGA or GPU to achieve both high accuracy and low power. With Xilinx and Nvidia’s sponsorship, contestants competed in two different categories: FPGA using Xilinx Ultra 96 and GPU using Nvidia TK2. All contestants used a large training dataset provided by DJI, a company renowned for drone technologies. The dataset contains over 14 GB of images with 100 different objects to detect. A hidden dataset was used to evaluate the performance of the designs in terms of accuracy and power. This year’s contest attracted over 110 teams from more than 10 countries/regions. In this session, the SDC organizers will first introduce the background and statistics for this year’s contest and announce the competition results. Then the top three teams in the FPGA and GPU category will present their designs.
|68.1||Introduction to 2019 DAC System Design Contest: Dataset, Statistics and Discoveries|
|Speaker:||Jingtong Hu - Univ. of Pittsburgh, Pittsburgh, PA
|Authors:||Jingtong Hu - Univ. of Pittsburgh, Pittsburgh, PA
Jeff Goeders - Brigham Young Univ., Provo, UT
Philip Brisk - Univ. of California, Riverside, CA
Yanzhi Wang - Northeastern Univ. , Boston, MA
Guojie Luo - Peking Univ., Beijing, China
|68.2||Presentation: GPU Category|
|Speaker:||Shouyi Yin - Tsinghua Univ.
|68.3||Presentation: GPU Category|
|Speaker:||Cheng Zhuo - Zhejiang Univ.
|68.4||SmartNet: A Bottom-up DNN Design to Unleash AI Capabilities for IoT|
|Speaker:||Xiaofan Zhang - Univ. of Illinois at Urbana-Champaign
|68.5||Presentation: FPGA Category|
|Speaker:||Cong Hao - Univ. of Illinois at Urbana-Champaign
|68.6||Presentation: FPGA Category|
|Speaker:||Kaan Kara - Eidgenössische Technische Hochschule Zürich
|68.7||Presentation: the FPGA Category|
|Speaker:||Boran Zhao - Xi'an Jiaotong Univ.