A Survey on Automatic Detection of Visual Bugs in Graphic Systems

PhD Qualifying Examination


Title: "A Survey on Automatic Detection of Visual Bugs in Graphic Systems"

by

Mr. Weiqi LU


Abstract:

Graphic systems, including renderers/frameworks like browsers, game engines, 
and visualization libraries, and applications such as web and Android 
applications, 2D/3D games, and GUI-based tools, are prone to visual bugs that 
compromise user experience and system reliability. These bugs manifest as 
interactive issues, where visual updates fail to align with user actions, or 
non-interactive issues, where displays deviate from code or user 
specifications. Manual inspection of such bugs is labor-intensive and 
inefficient, underscoring the need for automated detection methods. This 
survey investigates three main automatic testing methodologies: traditional 
approaches like differential and metamorphic testing, computer vision 
techniques including image processing and machine learning/deep learning 
models, and vision-language model methods leveraging prompting strategies. 
These methodologies shed light on the automated detection of visual bugs in 
data visualization (DataViz) libraries. Our empirical study reveals a heavy 
reliance on manual checks of the visual correctness in DataViz libraries, 
motivating future work to enhance the reliability of visualization through 
automatic visual testing approaches.


Date:                   Monday, 16 June 2025

Time:                   9:00am - 11:00am

Venue:                  Room 3494
                        Lifts 25/26

Committee Members:      Prof. Shing-Chi Cheung (Supervisor)
                        Dr. Shuai Wang (Chairperson)
                        Dr. Arpit Narechania