A survey of Image and Video Instance Segmentation using Deep Learning

PhD Qualifying Examination


Title: "A survey of Image and Video Instance Segmentation using Deep Learning"

by

Mr. Lei KE


Abstract:

Instance segmentation is a fundamental research topic in computer vision with 
many real-world applications, including image/video editing, scene 
understanding and robotic perception. Various instance segmentation algorithms 
have been developed in the literature with remarkable progress. However, their 
performance is still not desirable when deploying in complex real-world 
environments, such as segmenting highly-overlapping instances or objects of 
novel categories. Besides, effectively and efficiently leveraging the rich 
temporal information in video segmentation remains a challenge.

In this survey, we give a comprehensive review of deep learning-based 
image/video instance segmentation methods. We first introduce how to advance 
segmentation performance with the bilayer structure decoupling and 
commonality-parsing techniques. We then present an effective and efficient 
prototypical cross-attention network for video instance segmentation. Next, we 
further show an application of video instance segmentation: object-based video 
inpainting. Their limitations are also analyzed. In the end, we conclude this 
survey by summarizing several future research directions.


Date:			Wednesday, 6 October 2021

Time:                  	3:00pm - 5:00pm

Zoom meeting:
https://hkust.zoom.us/j/93090822445?pwd=YVY3Wm1mUVNadmV6RzluamJFTk5PQT09

Committee Members:	Prof. Chi-Keung Tang (Supervisor)
 			Dr. Dan Xu (Chairperson)
 			Dr. Qifeng Chen
 			Prof. Yu-Wing Tai


**** ALL are Welcome ****