A SURVEY ON FACADE SEGMENTATION, PARSING AND MODELING

PhD Qualifying Examination


Title: "A SURVEY ON FACADE SEGMENTATION, PARSING AND MODELING"

by

Mr. Chao YANG


Abstract:

Semantic parsing of facade has many potential applications. This paper 
surveys state-of-art techniques in facade segmentation, parsing and 
modelling. Most of the works assume regular structures on facade layout. 
Some of the methods are image-based and are model-free, by using the 
bottom-up image information alone to discover structures and patterns. 
Meanwhile only weak prior such as tiles, alignment, same-height rule are 
used. Other works are grammar-based, which shares the same framework of 2D 
shape grammar inference that encodes local and global features. The 
difference is that the formulations are slightly altered, as well as the 
inferring methods used change from one another. We address these two 
approaches separately and conclude with discussions on whether there 
really exists a procedural logic to facade parsing, and what are the most 
discriminative features that tell one city apart from another. At the end 
of the survey we present our own work of building city-scale facade 
database. We also show our novel approaches, based on our own database, 
that well handled the facade parsing and place recognition problem.


Date:			Monday, 9 September 2013

Time:                   2:00pm - 4:00pm

Venue:                  Room 3401
                         lifts 17/18

Committee Members:	Prof. Long Quan (Supervisor)
 			Prof. Chiew-Lan Tai (Chairperson)
 			Dr. Huamin Qu
 			Prof. Chi-Keung Tang


**** ALL are Welcome ****