I'm talking about complex structures , and especially (car, human face, plane)as first step in my project.
- Benefit to the blind is a good one. A program that could reliably, automatically describe an image in human form would be brilliantly useful.
- I fail to see how it'd benefit 3D designers and architects? I'm pretty sure they'd prefer to work with human (and most likely their own) descriptions of images!
- Image searching is the obvious one I can think of if it was done accurately, then searches could be based on the content of the image alone instead of the text around them.
- Filtering out suspicious material from web pages (if this was a reliable and fast algorithm browsers could do it on the fly and suspect images could be blocked rather than maintaining a blacklist of web pages.
- Data mining (automatic categorisation of images in a database and similar)