Hi friends,
Recently we launched a new software tool which helps content owners to organize their content in a very easy and sophisticated way.
With our technology you are able to find content you wouldn?t find using standard text-based search. Our software, Impala, adds visually-based tags to their content ? and it does so fully automatically.
As a software engine that you integrate into your own system Impala is tailored to search in digital images and in digital videos. The engine creates a pipeline with visual data flowing in on one side and the classification results flowing out the other.
Check out the movie to experience yourself how we arrange this for mainstream organizations like law enforcement agencies to search specific not by law approved content.
vimeo . com / 53593070
Some features.
- Tags images automatically, based on their visual content
- Enriches already existing tags (e.g. time stamps and geo-location)
- Further improves text search results for image and video search engines
- Available under web service agreement and available under software license agreement
- Ability to tune the recognition capacities to your company?s specific needs
- Maximal recognition accuracy: Evaluated at international competitions as TRECvid,
- ImageCLEF, and ImageNet where top performance is achieved over the years
The software tool contains a unique algorithm and can be used for images and movies.
The technology works straightforwardly for images. For videos, the input data is first segmented into shots, after which visual concepts are computed for each shot individually.
The engines can recognize a series of ?concepts?. Example concepts are: Airplanes flying, animals, guns, invoices, boats and ships, demonstrations or protests, female faces, flowers, and so on.
This is not the limit of what can be be recognized; the engine can also be customized to your specific needs. Manually selected training data is fed into the machinery, after which the software calculates new models. These newly created models are then used for tailored classification.
The training requires significant computing power. To speed up the process, we perform the training for you on a (large) grid of computers.
Would you like to have more information, you know where to find me.
Marc Oh