A Survey on learning based depth estimation methods

PhD Qualifying Examination


Title: "A Survey on learning based depth estimation methods"

by

Mr. Jiaxin XIE


Abstract:

Depth information is important for understanding 3D structure of the 
scene, so it plays an important role in computer vision and artificial 
intelligence. For example, depth map can provide additional signals for 
navigation and obstacle avoidance for unmanned systems. Also it 
contributes to augmented reality applications, in which it can help to 
inserting objects accurately on the space. This paper surveys various 
learning-based depth estimation methods and presents them in a common 
taxonomy. First, several monocular depth estimation methods are reviewed, 
which has developed rapidly with the popularity of deep learning. Next, 
advances in stereo methods are analysed. Moreover, we summary many other 
setting methods such as depth from multiple images, depth fusion from 
images and other sensor, depth from focus or defocus, depth from dual 
pixels and so on. Finally, we list some potential research direction.


Date:			Monday, 10 August 2020

Time:                  	4:00pm - 6:00pm

Zoom Meeting:		https://hkust.zoom.us/j/93285560352

Committee Members:	Dr. Qifeng Chen (Supervisor)
  			Prof. Pedro Sander (Chairperson)
 			Prof. Chiew-Lan Tai
 			Dr. Ming Liu


**** ALL are Welcome ****