The problem
Blind users meet images with no alt text, and machine captions can be wrong.
What to build
Build a tool that proposes descriptions and lets volunteers refine them, improving over time.
Success looks like
Descriptions rated useful by blind testers and a quick correction flow that feeds improvements.
Suggested technologies
- Vision-language models
- Human-in-the-loop
- Browser extension
- Quality scoring
- Crowd refinement
An original, independent statement. It is not an official Smart India Hackathon problem statement, and Arventiq Labs is not affiliated with or endorsed by that programme. The scenario is illustrative and does not describe any specific organisation. Free to use with credit.