Representation learning for unordered point cloud data

Szemenyei Márton
Kölső konzuelns:
Önálló laboratórium - Irányítórendszerek ágazat, BSc Vill.
Önálló laboratórium 1 - Irányítórendszerek főspecializáció, MSc Vill.
Önálló laboratórium 2 - Irányítórendszerek főspecializáció, MSc Vill.
Hallgatói létszám:
Szakdolgozat / Diplomaterv
Time-of-flight sensors (e.g. LiDAR) are widely used for detecting the distance of objects in the
surrounding. Data collected from such sensors is represented as a cloud of unordered points in the 3-
D space. While the use of convolutional neural networks is an intuitive choice for computer vision
problems where the input is represented as a 2-D matrix of pixels, it is still an open area of research
how to represent point cloud data for deep neural networks efficiently, also accounting for rotational
and translational invariance. Within this subject we explore state-of-the-art representation learning
methods with properties that enable their use for autonomous driving applications.