Enhancing ad blocker quality assurance using AI
summary
Opera builds locked split-screen on/off screenshots (locked viewport, lazy-load hooks) then runs visual validation models that compute an ad density score using image transforms and OCR to mark ad frames and promoted labels. A crawler checks top-50 sites per region every few days while user reports trigger focused runs; pipelines in GitLab raise tickets when density deltas jump and tests run via Playwright/Selenium against a private cluster. Models include open-source vision and LLM prompts.
AI-written summary · about
chapters (10)
- 0:00Introduction to AI in QA
- 1:49Limitations of reactive testing
- 3:39Proactive testing with AI
- 5:50Visual validation with AI
- 8:00Reliable screenshot creation
- 9:23Ad density scoring system
- 11:12AI-powered visual logs
- 12:39AI for test dataset generation
- 14:28Key takeaway and conclusion
- 15:14Q&A session begins
related talks
23mOperwatir and Testing at Opera
12mWhy am I seeing this: Democratizing End User Auditing for Online Content Recommendations
44mPanel Discussion | The AI Revolution in Testing: Opportunities and Challenges
14mAutomated Visual Regression and Accessibility Testing Made Easy
26mData, AI and design
42mHow Visual AI Makes Testing Easy
35mAutomated Accessibility Testing
24mAutomated Visual Regression & Accessibility Testing