Learning-based 3D Object Localization and Pose Estimation: A Survey

PhD Qualifying Examination


Title: "Learning-based 3D Object Localization and Pose Estimation: A Survey"

by

Mr. Yisheng HE


Abstract:

3D object localization and pose estimation targets to determine the 3D
spatial location and orientation of object instances in captured sensor
data, which plays a significant role in numerous real-world applications,
such as robotic manipulation, autonomous driving and augmented reality.
According to different problem formulation, it can be further divided into
two tasks: 3D object localization, also named 3D object detection and
object 6D pose estimation. Given sensor data of a scene, 3D object
detection returns the spatial location and extent of each object instance
via a 3D bounding box and object 6D pose estimation outputs the precise 3D
location and orientation of known objects. With the explosive growth of
deep learning techniques in recent years, many learning-based approaches
are introduced into this field and some major improvements have been
achieved. In this survey, we first give a brief review of the deep
learning techniques for automatically learning of object feature
representations. Then, we introduce the recent achievements of
learning-based 3D object localization and 6D pose estimation respectively.
Finally, we conclude some limitations and open problems in current
approaches and discuss possible future research directions.


Date:                   Friday, 14 February 2020

Time:                   3:30pm - 5:30pm

Zoom Meeting:           https://hkust.zoom.com.cn/j/561103463

Committee Members:      Prof. Long Quan (Supervisor)
                        Dr. Xiaojuan Ma (Chairperson)
                        Dr. Qifeng Chen
                        Prof. Chiew-Lan Tai


**** ALL are Welcome ****