Showcase: Coko


Deep Learning for image description

Coko makes it possible to recognize different vehicle parts with a neural network and to classify them according to their position in the vehicle.

In the field of corrosion research, vehicle parts have been photographed at different exposures and from different perspectives. This has resulted in more than 70,000 photos.In order to analyse the corrosion cases, it was necessary to classify the images according to the different car parts depicted The manual effort was about 300 hours. Our neural network in Coko can recognize these images with an accuracy of 83 percent and classify them immediately. Coko is continuously improved by control and possible corrections from users.

Possible applications: Automated quality control, logistics optimization, etc.