“DullDream” reverses the expected functions of a convolutional neural network (CNN). Instead of amplifying and highlighting the characteristics of an image to recognise it for categorisation in classes the “DullDream” reduces the specific characteristics in an uploaded image. The work name refers to Googles DeepDream, however, the effect is reverse. Users can upload images through the website http://dulldream.xyz and a"dull" version is generated. Besides the final generated image, in an advanced view, the user can see various steps of how the segments were isolated and the image generated.The original and generated image are displayed side by side in the Gallery of recent images http://v2.dulldream.xyz/gallery and when agreed upon also in an exhibition installation.
Description (in English)
Situation machine vision is used in