Xerox has extensive expertise in computer vision and imaging and this site presents some of our technologies through a range of easy to use demos. In many cases we use a combination of methods to solve real-world digital information management problems ranging from stock photography, digital asset management and personalised imaging through to cross-media search and smarter document technologies. We include here some mature technologies, and also some experimental demos which are in beta (or even alpha!) versions.
New ideas and demos will come and go on the page, as we try out new things and get user feedback. Take a look, have fun, and let us know what you think!
For many applications, you need to search for similar images based on an image query -- rather than a keyword search on the web which returns images from related web pages that contain those keywords. Our technology focussed on situations where there is a need to search large collections of high quality photos, digital assets, graphic desin elements, scanned documents or other media. Not everyone has access to a cloud, so the technology was developed around efficient compressed signatures, that work with all types of diverse media. The result is highly efficient, accurate retrieval that can search several million images on a single core in less than a second.
This demo also allows you to retrieve images based on the output of our image classifier, which you can try out
in more detail in the demo below!
In this demo we give some examples of image categorisation. This example uses training data to train a categorizer which recognises 706 different categories of images. For now it is trained to recognize 58 different types of sports, 136 species of animals, 111 species of plants, 237 objects seen in daily life, 41 famous monuments, and much more. You can vary the number of returned categories based on the precision you want. We have applied such technology to digital asset management and graphic design. It is also is part of our Smarter Document Management Technology platform.
We are committed to making great color accessible to everyone. Our research team developed an innovative technology for editing color, called Natural Language Color. You can now use common color words and phrases to change and improve images. Simple instructions like "Make the sky blues a lot more vivid" are translated by the software into complex color adjustments.
Realtors can improve the colors in their sales flyers. Small business owners can enhance products and imagery in their collaterals. Great color is now in everyone's hands. Try it out and let us know what you think.
Simple Personalized Imaging lets you create customized photos for birthdays and holidays, in much less time than premium tools like Photoshop. Xerox researchers have developed an easy-to-use tool that allows you to personalize your own photos by incorporating a text message in a natural way. You can easily insert new text or replace existing text in the image.
The software performs complex image analysis and 3D mappings to determine the perspective with which to place the text into the image. You can choose font, color and other special effects to make the text look even more natural.
A fun application you can share with your friends. Find out what your Facebook photos say about you! Catepix uses some of the image categorization technologies to profile you from your photos. Are you really as sporty as you claim? Explore your images with Catepix and share the results with your friends.
Many methods for image classification are based on recognition of parts -- if you find some wheels and a road, then the picture is more likely to contain a car than a giraffe. But what about quality? What is it about a picture of a building or a flower or a person that makes the image stand out from the hundreds which are taken with a digital camera every day? Here we tackle the difficult task of trying to learn automatically what makes an image special, and makes photo enthusiasts mark it as high quality.
Experiment with our newest application (still in the alpha stages!) which retrieves a group of images from a class, and then tries to predict which ones are normal, and which ones are of high quality. Do you agree with the system?