Open Peer Review. Open Publishing. Open Access. Open Discussion. Open Directory. Open Recommendations. Open API. Open Source.
A LiDAR Point Cloud Generator: from a Virtual World to Autonomous Driving
Xiangyu Yue, Bichen Wu, Kurt Keutzer, Alberto Sangiovanni-Vincentelli, Sanjit A. Seshia
Oct 31, 2017 (modified: Oct 31, 2017)NIPS 2017 Workshop MLITS Submissionreaders: everyone
Abstract:3D LiDAR scanners are playing an increasingly important role in autonomous driving as they can generate depth information of the environment. However, creating large 3D LiDAR point-cloud datasets with point-level labels requires a significant amount of manual annotation. This jeopardizes the efficient development of supervised deep learning algorithms.
We present a framework to rapidly create point clouds with accurate point-level labels from a computer game. The framework supports data collection from both autonomous-driving scenes and user-configured scenes. Point clouds from auto-driving scenes can be used as training data for deep learning algorithms.
We show a significant improvement in accuracy (+9\%) in point cloud segmentation by augmenting the training dataset with the generated synthesized data.
Point clouds from user-configured scenes can then be used to systematically test and analyze neural networks by sampling scenes in the scene modification space. We also propose a method to do automatic calibration between the point cloud and captured scene image.