Computer Vision News - April 2016

Application Clarifai COMPUTER VISION NEWS Clarifai ’s visual recognition solution was born in 2013, when its founder Matthew Zeiler decided to build up on his previous works and studies to create a service that would be able to recognize things in images. That solution was perfectly able to classify all photos on his own computer, recognizing all the pictures of his dog, his favourite sports team and so on. 4 If this can be very useful to search on one’s own PC, how useful can it be for an enterprise looking to search over millions and billions of photos? This was the marketing opportunity which brought this neural networks technology to the market under the umbrella of this new start-up. The very first task of Clarify was to try its luck at a computer vision competition called ImageNet . Clarify ended up winning all the five top places, which was a formidable kick-off for the new venture: this machine learning solution had obviously a case and was due to have an impact on a very high number of users worldwide. 30 months later, Clarifai has been selling his products through a high variety of different vertical markets. Anybody can check the demo ( try it here ) on his on her own photos and see how predicted tags are offered in no time, together with a choice of similar images. Not only objects are identified, but also descriptive words like love and romance are automatically predicted. An additional functionality allows, given a specific image, to find similar images. Similarity is found at a much deeper level than just template or colour matching. The application literally understands the image and this capability enables it to automatically identify similar content. “ This application literally understands the image and automatically identifies similar content ”