Recognizing recurrent elements of Martin’s drawings with a deep neural network.


In their new collaboration, artist Shantell Martin and computational cognitive neuroscientist Sarah Schwettmann set out to re-humanize the concept of ‘process’ and uncover elements of artistic identity within. This project began with an intent to understand the consistency underlying the stream-of-consciousness outpouring that is Martin’s drawing, evolved into an effort to understand the relationship of that consistency to artistic identity, and moves forward with questions of the uniqueness of that identity.


Over the course of three months, Martin completed a series of 300 drawings developed from 100 initial hand drawn templates. Each template, a continuous line serving as the foundation of a future drawing, was sent to Martin three times, weeks apart, to be completed in her characteristic style. After recognizing consistency in the artistic choices Martin unknowingly made to complete each set of three drawings, Schwettmann and colleagues at MIT set out to see if they could define deeper structure underlying the visual similarity of different drawings.

To do this, they trained a deep neural network to recognize recurrent elements of Martin’s 300 drawings, potentially enabling it to learn her artistic style. After training, if the deep network can successfully predict how Martin would complete a given drawing, we can suggest that it has identified key elements of her identity in the artistic



But where does this leave the artist? A common fear is that new technologies will ultimately replace the artist, in an era where resources are increasingly being diverted from the arts. Alternatively, the advent of technology as a window into identity could further the potential of art as a tool for self-expression, as awareness of one’s artistic process, and ultimately oneself, can powerfully refine and evolve artistic style.  The 300 line drawings will be on exhibit in the offices of FiftyThree, makers of Paper, this November. Visitors to the opening of the exhibition will have the chance to visualize their individual engagement with Martin’s artwork using mobile eye-trackers, and also experiment hands-on with their own ability to learn Martin’s artistic style.