Scene text access: a comparison of mobile OCR modalities for blind users
summary
Study compares three mobile OCR workflows with seven blind participants: SeeingAI Short Text (live OCR, fails while camera moves), Spot+OCR (server-based FCN text spotter at ~2 fps that vibrates to prompt a stable high-res photo, then OCR; spotter precision ~64%), and Guided OCR (guided 360° high-resolution shots fed to Google OCR). Guided OCR yields the highest recall in open spaces; Spot+OCR improves detection over SeeingAI unless users perform tactile exploration; SeeingAI is faster but often misses signs when users cannot frame text.
AI-written summary · about
chapters (17)
- 0:00Introduction to research
- 0:26Scene text access challenges
- 1:09Problem with camera-based OCR
- 3:18Text spotter solution explained
- 4:42Spot+OCR system design
- 5:45Guided OCR for open spaces
- 6:48Experiment setup details
- 7:34Participant testing results
- 10:39Performance comparison analysis
- 12:03User feedback and preferences
- 13:05System limitations and trade-offs
- 14:30Future interface improvements
- 14:52Q&A session begins
- 15:12Question on panoramic OCR
- 16:39Question on RFID alternatives
- 18:11Question on language model integration
- 18:52Final Q&A wrap-up
related talks
37mCan you Read This? How I Improved Text Readability Online for Visually Impaired
29mCan you Read This? (Or: how I Improved Text Readability on the Web for the Visually...
44mOCR Systems: Bringing Accessibility Where It’s Missing
09mSmart Glasses for the blind
21mA Face Recognition Application for People with Visual Impairments: Understanding Use Beyond the Lab
27mOne blind weekend
45mBe My Eyes - Leveraging Generative AI to help the blind and visually impaired people
15mStreetViewAI: Making Street View Accessible Using Context-Aware Multimodal AI