Learning what a wind turbine is

Since 2015, machines can dream and their dreams, like the one below, looks like the image below.

Behind these new images, a new type of algorithm : deep learning. Since then every single media on the planet has talked about deep learning and its applications.

Deep Learning is a specific Machine Learning algorithm. Most times the scientific community is looking for ways to modelise a problem to be able to predict what will happen in the future. Machine learning experts try to build a machine able to solve a problem by itself (basicaly build a model by itself). The machine has a lot of parameters to set to be able to adapt to a given problem.

We all imagine a machine at school learning bit by bit but most of the time, it’s way more brutal : if you want your machine to be able to recognize a blade, you will have to give it thousands if not millions of examples of blades.

With these examples, the machine will propose answers and we will tell if it’s good or not so it can adapt.

If you repeat the operation long enough and if your database is a good sample of your problem, you can manage to make the machine very efficient. For example, at Cornis, we are now able to describe blade images and automatically detect main elements as represented in the figure below.

Such a description of a blade is used everyday in our algorithms to automatically detect defects, accelerate expertise and to automatize stitching.