Caption Crawler: Enabling Reusable Alternative Text Descriptions using Reverse Image Search
summary
A browser extension plus an Azure cloud service uses a reverse image search API and a multi‑threaded crawler to fetch captions found elsewhere on the web and stream them back to the page. An exhaustive crawl of 481 popular pages finds over 8,000 images lacking alt text and captions about 12% of them (~1,000 images) with an average latency of 18 seconds. A user study with 14 screen reader users shows reduced uncaptioned images and mixed caption quality; limitations include latency, encrypted images, and inconsistent language/detail.
AI-written summary · about
chapters (8)
- 0:00Introduction to Caption Crawler
- 1:55Methods for generating image descriptions
- 3:39How Caption Crawler works
- 7:12System performance evaluation
- 8:38User study with screen reader users
- 11:31Implications for search engines
- 12:15Conclusion and future work
- 12:49Q&A session
related talks
22mFAULT TEXT: The Adversities in Automated Alternative Text with Henri Helvetica
01h 00mAI and alt text
50mWhat is “equivalent” text alternatives for artistic and cultural visual content
51mCaptions and Beyond Scaling Media Accessibility Audits for Title II Compliance
16mThe positive alt-itude! An inclusion tool for your accessibility
55mBridging the Digital Divide: Empowering Accessibility Through AI Innovations
34mThe power of alt and lang attributes
48mText-To-Image Accessibility