January 30, 2020, 13:30 - 14:00 @ Y25 (Room: Y25-H-38)
Deep Learning for Feature Extraction from Point Clouds
Dmitry Kudinov, Senior Principal Data Scientist at Esri Inc., USA
Summary: LiDAR sensors are an important and reach source of high-precision 3D point clouds, which are quite useful in modern urban design and planning. Such point clouds are easy to collect but slow and expensive to segment into classes and objects. We are going to talk about various applications of common deep neural networks in segmenting and labeling point clouds for fast reconstruction of 3D building models at scale, detection of overhead conductors and utility poles, street furniture etc. In conclusion, we are going to talk about how deep learning can be used to segment continuous meshes produced by photogrammetric processes.
Bio: Dmitry Kudinov is a Senior Principal Data Scientist at Esri Inc. His expertise lies in Geospatial Machine and Deep Learning, particularly in transportation and remote sensing; segmentation of LiDAR point clouds and spatial feature extraction, 3D content generation and simulation technology for autonomous agents. He received degrees of M.Sc. in Applied Mathematics as well as M.Sc. in Computer Sciences. He is also a Self-Driving Car Engineer for Autonomous Vehicles.