Vision & Deep learning

Interpreting images is so deceivingly easy and natural for humans that we do not fully realize how difficult vision is. For computers it is not obvious at all how to reconstruct space and time around us, and how to recognize objects from pixel data. Deep learning, in particular convolutional neural networks (CNN’s), is a recent breakthrough technology inspired by biology that aims at solving (part of) this problem. Understanding vision helps to comprehend and appreciate the power and potential of deep learning algorithms. Solving vision enables a robot to navigate and interact with real-world objects. I will give an update on the latest developments and achievements of deep learning with respect to vision and discuss both the similarities and the differences between human vision and artificial vision. I will also discuss today’s possibilities of programming and/or applying deep learning with software frameworks and API’s yourself. Democratisation of deep learning has created tremendous opportunities for innovations. New AI applications are within reach for every software engineer.