Cognitive Workbench Visual Solutions

Visual Solutions

Computers and machines can see through cameras, but they hardly perceive their visual environment. So they are more or less blind (and even deaf). ExB uses autonomous deep learning algorithms to identify objects and structures in pictures so that computers can learn to see. What seems to be easy for humans is a hard task for AIs.

How deep learning works

Deep Learning is based on artificial neural networks, which roughly mimic the human brain  through the  simulation of a tightly woven network of simple neurons. These neurons are divided in multiple levels. In a picture recognition task for example the first units register only brightness values of pixels. The next level would recognize that some pixels are connected and represent edges. Subsequent levels decide between horizontal and vertical lines and so on until the neurons could identify an object.


Deep learning powered systems can top human accuracy in recognition tasks


Not only deep learning algorithms can handle Big data they even get better with more data


Deep learning approaches are extremely flexible, new domains can be acquired with ease


Image recognition is stunning fast, thousands of images can be processed in seconds

What a seeing computer can be used for

Object recognition is important for machines like autonomous cars, they have to recognize lane markings, traffic signs and other obstacles in their way. But computer vision can also be used for other tasks, for example face recognition (finding, marking and tagging), abnormalities in tissues or finding logos in social media.