Advanced features and challenges from the Labs
Seach by Drawing
When we were researchers at the University, we have been working hard on inventing new ways for searching content, through visual means and browsers, but also through alternate modes for querying. Of course, it's important and possible to search by keywords or by similarity through lateral search. But we are proud to also power "search by drawing". This feature has been test-deployed in several venues, and we will soon be ready to offer it to a wider audience through our cloud platform.
Enhancement, Editing and Generation
Computational crativity tools using AI will offer new means for people and machines to interact and produce new ideas, aesthetics and content. We research on photorealistic picture generation and manipulation. We don't plan to replace the photographer by a machine anytime soon, but are developing techniques that will enhance photographic or vidéo content, enabling archives to do more with their heritage catalogs. We plan to offer increased image resolution and recolorization functions through our cloud platform. More generally, we will reverse the effect of aging and degraded copies of analog support.
TEXT DETECTED: "MERRY CHRISTMAS NEW YE"
Images and words
Optical character recognition extracts important information by transcribing captions and titles within video content, further extending the reach.
Novel AIs can do remarkable things. But they can also go wrong sometimes. Cultural bias is one such example. We want to fix that. We are working with people with various backgrounds and a technology platform that enable continuous improvement of our computational models for visual understanding towards a better account of cultures, regions, and interests.