How can a computer “see”? For an artificial intelligence to recognise colours, brightness and shapes it needs not only training but also various filters, so called Convolutional Filters. These form a Convolutional Neural Network (CNN) and discern different elements of an image. The CNN in our Understanding AI exhibition has learnt with a lot training data to detect certain objects. The artificial intelligence is however not always flawless – often it can be deceived or makes mistakes in image recognition that humans would not make. In this Home Delivery segment we show how this artificial intelligence works, what’s necessary for image recognition, and what can go wrong.
