AI Systems for Accessibility
summary
Deque shows production AI systems that combine a computer vision model that identifies UI elements from pixels with large language models that generate tailored guidance and code fixes. The pipeline compares visual detections to markup to cut false positives, surfaces axe‑core issues, and suggests both semantic and minimal HTML remediations (preserving class names). Suggestions are validated against the axe API and a secondary model, link to Deque University citations, and ship as AXE DevTools Pro features and a preview Access Advisor.
AI-written summary · about
chapters (10)
- 0:00Introduction to AI systems
- 2:00Computer vision advancements
- 4:37Large language model basics
- 7:13Contextual AI capabilities
- 12:21Prompting and adaptability
- 15:59Challenges and human control
- 20:56The accessibility challenge
- 23:56Introducing Axe Advisor
- 28:18AI remediation tool demo
- 39:04Question and answer session
related talks
50mEnhancing Accessibility with AI and ML
28mAutomated Accessibility Testing
52mUpdate on The State of Accessibility
19mAgile Accessible Software Development
34mAdding accessibility into Unit Tests
34mAI for Accessibility: Empowering Inclusive Digital Experiences
30mTesting for Accessibility with aXe
43mHow to Use AI to Build Accessible Products