SoAIR2019

Technical Presentation

---

• Zhiliu Yang, Perceptin, Singapore

Title: DragonFly: A Localization and Perception Module for Reliable and Affordable Autonomous Driving

Abstract : Autonomous Driving technologies have attracted enormous interests from both academia and industry in last few years. By equipping with multiple-ways Lidar, complicated tasks can be handled by autonomous vehicles. But the expensive cost of Lidar module prevents the popularization of this solution. Based on this, we propose an affordable method by using our DragonFly module to fulfill autonomous driving tasks. In this talk, a thoughtful introduction for DragonFly module will be given, it is comprised by 4 HD (1080P) global-shutter cameras (a stereo pair in the front, and a stereo pair in the back), an IMU device, a Nvidia Jetson TX1 computing submodule and a rich set of I/O. It provides three functions for autonomous driving: perception, localization, and panoramic video streaming and maintain a decent performance. Several demos will be shown as well, such as localization via a VIO SLAM algorithm, perception made up by real time road conditions, vehicle and pedestrian detection, spatial information of environment, depth information of vehicles and pedestrians.

Bio: Zhiliu Yang, is currently a Researcher at PerceptIn. He receive his B.E. in Electrical Engineering at Yunnan University and M.S. in Electrical Engineering at Beijing Institute of Technology. He is interested in Computer vision, Deep Learning, and High Performance Computing.