Researchers have created an artificial intelligence software that uses photos to zero in on documents from the internet with greater accuracy than ever before.
The new system, developed by researchers at Dartmouth College in the US, uses pixel data in images and video - rather than just text - to locate documents.
By studying results from text-based image search engines, the software recognises the pixels associated with a search phrase and applies them to other photos without tags or captions, locating them more accurately.
"Existing popular search engines, strip away information contained in photos and use exclusively the text of Web pages to retrieve documents," said co-author Lorenzo Torresani from Dartmouth College.
New research shows that "modern machine vision systems are accurate and efficient enough to make effective use of the information contained in image pixels to improve document search," he added.
The new software improved search precision by 30% compared to purely text based searches.
The findings appear in the journal PAMI (IEEE Transactions on Pattern Analysis and Machine Intelligence).