The IViSE workshop was part of the prestigious Computer Vision and Pattern Recognition (CVPR) conference held in Nashville (USA) in June 2025.

Image search tool

Exquisitor is the result of 5 years of research by Ujjwal Sharma and Stevan Rudinac and Omar Shahbaz Khan and Björn Þór Jónsson (Reykjavik University). The system has been tested on lifelog images, underwater footage, and medical videos, and has been rigorously evaluated at international competitions

Finding the right video or image in a huge archive can feel impossible. This interactive system combines conversational search with feedback, helping users quickly locate specific content - even in collections with millions of images or thousands of hours of video.

Fast and accurate

Exquisitor retrieves results in less than a second, regardless of collection size. On datasets with over a million video segments, it consistently shows relevant results in the top 10.

Real-world applications

Exquisitor is more than a research project—it offers practical solutions across many domains:

• Personal media: quickly find moments in private video collections or photo libraries.
• Healthcare: support patients with memory conditions, such as Alzheimer’s, by helping them access and recall visual memories.
• Specialised datasets: distinguish between visually similar underwater footage or complex medical procedures where text-based search fails.
• Research and education: provide academics and students with new ways to explore and analyse large-scale multimedia archives.
• Cultural heritage: make vast collections, such as those at national archives and museums, more accessible to the public.

What’s next

The team plans to release Exquisitor as open-source software, making it a powerful tool for researchers and general users worldwide.