Can you Read This? (Or: how I Improved Text Readability on the Web for the Visually...
summary
Builds a practical system that flags whether text over images is actually readable by people with vision loss rather than just naming the topic. Uses OCR experiments (Tesseract fails on 72 dpi web screenshots; EasyOCR helps), then extracts characters with CRAFT character heatmaps and runs a binary per-character classifier, counts poor characters and thresholds the text. Trained on mixed objective/plain-background labels, integrated into a Site Scanner product, finds many failures especially in travel/news/art sites; cites 250M+ visually impaired (217M severe, 36M blind).
AI-written summary · about
chapters (4)
related talks
37mCan you Read This? How I Improved Text Readability Online for Visually Impaired
51mReadability & Web: Let's build great inclusive projects
50mYou can't read this sentence
27mCrafting Comics for Literally Everyone
48mTypography, Readability and Digital Accessibility
18mReadability & Design Building inclusive digital experiences
56mTypographer vs. Accessibility
25mYou can’t read this sentence - A11y automation