The European Classes, Euronet 24 October - 11 November 2018

Accordion

Among the ever evolving machine learning technologies, so-called ‘convolutional networks’ (ConvNets) are increasingly developed to recognise objects within photographic images.  The artists have retrained these image recognition networks to include European artefacts, creating an image dataset. The underlying questions is how Europe’s diverse cultural output can be represented within this dataset. And what is a networks’ capacity to recognise what is deemed European. By illustrating the networks’ ability to draw out cultural bias, the artists show a networks understanding and interpretation of Europe in 2017.

The video problematises a machine’s visual understanding of European culture in light of the geopolitics of data sets. It questions the networks capacity to learn on the basis the collected images. Visitors can access the EuroNet via the touchpad.

On the technical side, the image dataset was collected using a combination of manually and semi-automated image gathering techniques. More popular categories, such as Whiskey, were easier to obtain using semi-automated image scraping with imsearch-tools, while others were manually pieced together. The bounding boxes that define the objects within the images were all manually annotated using both in-house and crowdsourced labor. In total there are over 20k images in the dataset, which can all be downloaded as JSON files with URLs and bounding boxes from the project website at ImageNet.xyz. Each class contains approximately 50-100 images.

About the artists:

Constant Dullaart‘s practice reflects on the broad cultural and social effects of communication and image processing technologies, from performatively distributing artificial social capital on social media to completing a staff-pick Kickstarter campaign for a hardware start-up called Dulltech™. His work includes websites, performances, routers, installations, startups, armies, and manipulated found images, frequently juxtaposing or consolidating technically dichotomized presentation realms. Recent solo Exhibitions include Cultural Matter, LIMA, Amsterdam; 100,000 Followers for Everyone, FOAM, Amsterdam (2018); Phantom Love, Up Projects, London (2017); Synthesising the Preferred Inputs, Future Gallery, Berlin; Deep Epoch, Upstream Gallery, Amsterdam (2016); The Possibility of an Army, Kunsthalle Schirn, Frankfurt; Jennifer in Paradise, Futura, Prague; The Censored Internet, Aksioma, Ljubljana (2015); High Retention Slow Delivery, Jeu De Paume, Paris and Stringendo, Vanishing Mediators at Carroll / Fletcher, London; (2014). Group Exhibitions include I Was Raised on the Internet, MCA, Chicago; When Facts Don’t Matter Lismore Castle Arts, Lismore; (2018); Open codes, ZKM, Karlsruhe (2017-2019); Collecting Europe, V&A Museum, London (2017), Electronic Super Highway, Whitechapel, London; Final Goods, Kunstverein Hildesheim (2016);Then They Form Us, MCA, Santa Barbara; When I Give, I Give Myself, Van Gogh Museum, Amsterdam; Algorithmic Rubbish, Stedelijk Museum Bureau Amsterdam (2015); Casting a Wide net, Postmasters, NYC, USA; Online/Offline/Encoding Everyday Life, Transmediale, Berlin (2014). Dullaart has curated several exhibitions and lectured at universities and academies throughout Europe, most currently at the Werkplaats Typografie. Recently he has been awarded the Prix Net-Art 2015 and was a resident at the ISCP in New York in 2017.

Adam Harvey is an artist and researcher based in Berlin exploring societal impacts of networked data analysis technologies with a focus on computer vision and counter-surveillance. He is a graduate of the Interactive Telecommunications Program at New York University (2010) and previously studied engineering and photojournalism at the Pennsylvania State University.

This item is part of

Exhibition: Algorithmic Superstructures

Other items in this program