business triangle technical hexagon

How I won the Alibaba self-driving LIDAR point cloud segmentation competition

Technical talk | English

Theatre 25: Track 1

Thursday - 14.40 to 15.20 - Technical



In this talk I’ll explain how I competed against 1400+ participants to perform real-time segmentation of LIDAR point-clouds for Alibaba’s self-driving car research.

I will explain how I got both #1 prize for Alibaba’s competition along with most innovative solution on CCF 大数据与计算智能大赛(CCF Big Data & Computational Intelligence Contest简称“CCF BDCI”) held in Shenyang in China, along with tidbits of competing in China.

Self-driving cars will come in different flavours, and it is expected that commercial fleets (robo-taxis) will be equipped with LIDAR sensors. LIDAR sensors work by casting multiple laser beams and recording reflections around the car. These reflections need to be processed to build a computational representation on the surroundings of the car to detect obstacles, etc.

LIDAR perception is still an open problem and as such companies interested in solving it often set up competitions to advance the state of the art.

Alibaba is one company interested in self-driving cars and organized a competition held within the 6th CCF Big Data & Computational Intelligence Contest (CCF BDCI) to perform real-time segmentation of lidar point clouds acquired from Alibaba’s fleet of self-driving cars.