Cortexica - About Cortexica

At Cortexica, vision is very much at the heart of everything we do, from inspiration to aspiration, vision is at the core of our research, technology and commercial applications

Inspirational vision (research)
In 2002, in the Bioengineering labs of Imperial College London, Dr Anil Bharath and Dr Jeffrey Ng found inspiration from the way human beings see and have spent the last six or so years reverse engineering how the human visual system works.

Success was achieved in 2006 when Bharath and Ng effectively modelled how human neurons react to visual stimulus, accurately mimicking the Primary Visual Cortex, also known as V1, together with its indifference to variance of lighting conditions (e.g. bright or dark scenes), orientation (rotation in and out of plane) or geometric distortion (e.g. curvature).

Applied vision (technology)
In 2008, Cortexica Vision Systems Limited was formed; soon after, Bharath, Ng and team built a computerized cortical key-point based image recognition technology (vision system) upon which modules for different applications of image recognition (IR) can be placed to affect a product or service, such as looking for a logo in a stream of video, recognizing a human face or matching an image against a database of images.

As is often in business success, timing is important, and whilst Bharath and Ng were building a ground breaking image recognition system to run on general purpose processors (CPUs) the world of computer graphics hardware was evolving fast to provide twin, quad cores, then hundreds and now thousands of processing units, such that today’s commodity GPU-based hardware can combine to offer around one thousand processing units that operate in parallel which marries directly to how a human’s neurons, and in particular, how the visual neurons combine to provide extremely fast and efficient recognition of what the eyes are seeing.

Bharath, Ng and team have set about harnessing these massive computational leaps in GPUs technology and in inverting the traditional use of the GPU (to generate a complex image) they have built the world’s first real-time cortical-based key-point IR system that utilizes GPUs to process an image (rather than build one) to get unique image data out.

Future vision (aspiration)
We are proud to share with you our vision for the future – that future is Visual Search – where a user no longer has to describe something in textual terms, but simply has to take a picture of what they want to find.

In many ways, we have already started on that journey, specifically with WineFinder, where a user, anywhere in the world, can use a mobile device to search for an item (wine) using an image (photo). This is the first step on the road towards Visual Search.

We have started to map that road and as we build out applications of our IR for search for discrete domains, our technology will evolve, so too will our domain experience, such that we will be best placed to provide the launch pad for the next revolution in search, Visual Search.

Intelligent vision (commercialisation)
Having built a robust IR platform from which to springboard IR applications, Cortexica set about identifying commercial opportunities that would benefit from a reliable IR capability, especially where humans are currently relied upon for the processing and subsequent (subjective) interpretation of the image data.

BrandTrak for Video
BrandTrak for Video offers a fully automated empirical measurement of brand exposure within video, from low-grade online video with visual artifacts to broadcast television; from cinematic movies to Super HD and 3D - continual, quantitative and repeatable brand scoring using our proprietary C-impact.

The C-impact
The C-impact (pronounced See Impact) is exactly as it sounds - a value attributed to the impact a logo has in relation to the image (frame of video) within which it was identified.

C-impact is an equation of factors such as:
- Size (area taken up),
- Position (on screen),
- Occlusion (obscuring),
- Orientation (in/out of plane), and
- Salience (shape, colour and contrast)
- Temporal dynamics
- Velocity or acceleration
Baseline currency
And because the C-impact is calculated the same way, using the same factors for each brand instance on a per frame basis, it is repeatable, time after time, across all brands regardless of brand owner – in effect a brand exposure currency.

This enables C-impact to be used as a baseline reference point from which all subsequent brand instances (exposures) are measured and tracked against.

WineFinder
As its name hints, WineFinder is an application that enables the user the ability to find out more about wine. In fact WineFinder is a mobile (smartphone) application that enables discovery of wine to be fun and non-intimidating.

It opens up the otherwise esoteric world of wine by allowing the user to take a photo of a wine bottle label, then, using the same cortical key-point based IR technology (used within BrandTrak) have that wine identified.

Once correctly identified, information about that wine served back, along with the opportunity to rate it, share it (with their social network) and purchase it directly, or indeed an alternative offered up by a clever little reference engine.

Reference engine
The reference engine understands what appropriate alternative wine to offer up, based on such things as:
- The attributes of the wine, e.g. grape, colour, appellation (region), vintage, etc.
- Relevant promotions of the retail wine merchant.
- Friends (from their social network) recommendation.

The first incarnation of this revolutionary mobile IR application is being developed for the iPhone in partnership with wine data and image repository holders and retail wine merchants.