Launched this week, Imprezzeo is a search software company that aims to improve image search, offering a way to use images rather than text to search for pictures.
The technology behind Imprezzeo’s image search was developed by University of Wollongong and the University of Queensland in Australia, though the company s based in London.
We caught up with CEO Dermot Corrigan to find out more…
What does Imprezzeo do?
We’re a visual search company which specialises in helping people finding images using other images.
With keyword image search results can be unreliable – our technology offers a better way to refine that search; by choosing a sample image that closely matches the search as a visual clue and finding other similar images. Sample images can also be uploaded to find similar results.
The search engine will pick up characteristics of these images and use these to find related pictures.
We have two technologies that we use to find images: content-based image retrieval (CBIR) and facial recognition, a combination of these is used to make our results more accurate.
What applications does the technology have?
Initially, we are focusing on the B2B marketplace, targeting companies that have large numbers of digital images, like news agencies, photography firms and publishers. This approach should benefit us most in the short term. We are in the advanced stages of talks with some well known names.
Is there a version of the technology for anyone to try?
Our biggest challenge here is that we need a large amount of content to be able to showcase out image search. We hope to produce a beta version soon for people to see how our search works.
Can this be used to improve image search online?
This is something which can be used to complement existing search engines. Currently image search online could be much more effective, it is at a point where text search was five or ten years ago.
Image search at the moment is all about interpretation of the text associated with images, and the results are not always accurate.
Imprezzeo could be used as a way of refining current image search results, either by choosing the best results as a sample and finding matches, or by uploading pictures to find similar ones.
How does Imprezzeo’s technology differ from that developed by companies like Polar Rose?
The underlying principles are similar, though it seems Polar Rose are focusing on facial recognition, which is only part of what we do.
For facial recognition to work, it first needs examples or training sets so that it can identify similar faces. They are trying to tag as many faces as they can, and are taking the approach of getting the online community to do much of the legwork.
How does Imprezzeo solve this problem?
We need to do this as part of the set up process when working with any database of images. We will build training sets so that the system can learn to recognise the images we need to find.
The system will then automatically identify people in images, or flag up when it doesn’t.
How many people are in the team?
We have twelve in the Imprezzeo team at the moment, though we are scaling up the development side.
How are you funded?
We are backed by Independent News and Media, who are best known in the UK for publishing the newspaper.