Exploring the “Exactitudes” Portrait Series with Deep Learning Neural Networks (2015)
Deep learning neural networks have shown unprecedented results in concept identification in collections of images. For his graduation project, student Sam Verkoelen applied them to the collection of portraits from the "Exactitudes" artwork by Dutch artists Ari Versluis and Ellie Uyttenbroek. Through these portraits, the artists attempt to document and identify different subcultures in societies worldwide.
By applying such unsupervised machine learning techniques to the photo collection, Sam aimed to understand the relationship between portraits and series, uncover underlying features and dimensions of subculture imagery, and generate new images. As such, the work provides a useful case example of understanding structure that is uncovered by deep neural networks, as well as a tool to analyze the underlying structure of a collection of visual artworks, and a very first step towards a robot curator.