Leading Vision AI

Advancing the Frontiers of Object Recognition and Visual Search

Object Recognition

The ability to automatically identify objects, such as products or specific items, from within a photo image. Cortexica can segment and identify multiple complex items from within one image. Used for first identification reference, or as a start for a Visual Search.

Image Analysis

This approach analyses all objects, perceptual elements, such as settings, colours, patterns and shapes in the image, to provide a summary of the mood, theme or emotion of the image. This is useful for extracting insightful information from a simple image.

Visual Search

Cortexica’s patented technology replicates the human ability to find visually exact and similar matches to an image from a dataset of other images. This requires no training and can be applied to images of any content. Cortexica customises its suite of algorithms for each use case, replicating the most natural way of providing perfect search results.

Video Analysis

An extension of Image Analysis, Video Analysis provides revolutionary insights into continuous streams of vision, be it pre-recorded or in real time. With no manual tagging or manipulation, Cortexica’s VA approach opens up the world of interactive video in commerce, training, object and location identification and action recognition for security and health and safety.

With over a decade of investment and focus, Cortexica’s technology has matured into the best reverse image search solution in the world

In 2005, Cortexica’s patented technologies were conceived in the labs of Imperial College London by world renowned founder’s Dr. Anil Anthony Bharath and Dr. Jeffrey Ng.

Our dedicated and growing team of 35 R&D scientists, which includes 15 PhDs, continue to develop and enhance our algorithms and supporting architecture to maintain the most responsive vision AI software available.

Using our own code base means we can combine and adjust configurations to meet the demands of individual clients. This customisable approach is delivered using fully automated RESTful APIs that our clients can plug into their software, hardware, apps and websites.

AI – One Method Does Not Suit All

Combining our core technologies provides the custom solution you require

Computer Vision Methods

Our patented methods process several visual channels in parallel, such as scale, texture, colour and shape, looking at 1500 points of interest in each image. Depending on our client’s industry we can utilise other techniques. For example, within fashion we can use face/body detection to deliver higher accuracy.

Deep Learning

Depending on requirements we utilise different models (RNN & CNN) to learn, detect and categorise. We can even base our models on our clients’ own unique images and metadata, as we firmly believe that one size does not fit all. We then automatically identify and categorise any of our client’s image assets and auto tag their products.


Our technologies can automatically detect and segment the image, automatically apply the right configurations and process a response in under a second.