On the basis of the characteristic aspects of a picture, certain computers can tell us what the picture is showing. They can learn this in the same way that young children are able to learn to recognise images. Further improving these techniques opens the way to a whole range of new applications. This is a field in which we are carrying out much research work.
For years already, we have been trying to train computers in image recognition. By now, much more is possible, due to the development of convolutional neural networks. Through this kind of networks, enormous amounts of labelled images give constant feedback during an image recognition task. This makes it possible for the computer to learn faster.