Abstract
Identifying software regressions and user experience issues in complex applications from video recordings can present a challenge for certain quality assurance methods, which may be inefficient for analyzing long video sessions. The disclosed technology can use a large generative artificial intelligence model to perform a granular, for example, minute-by-minute, analysis of video recordings captured from an application's user interface. This system can process extended video streams to detect potential visual and functional anomalies, such as incorrect camera behavior, faulty user interface callouts, or inferred voice guidance errors. Use of this method may help to automate the identification and reporting of software regressions, which can contribute to an increased bug detection rate, more standardized reporting, and improved software release quality.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Cheng, Shu-Wei; Huang, Andy; Joshi, Megha; Borgnino, Juan; Sadasivam, Yoga; Ding, Oscar; Hua, Jessica; Li, Zening; Wang, Fei; and Kodur, Kamakshi, "Video Analysis for Software Regression Detection Using Generative Models", Technical Disclosure Commons, (September 23, 2025)
https://www.tdcommons.org/dpubs_series/8620