alsino skowronnek

The Amazing Augmented Tagger Machine

Graffiti Tags generated by Artificial Intelligence

01. The project
The "Amazing Augmented Tagger Machine" explores the potential of artificial intelligence in the creative process of graffiti writing. It employs machine learning (ML) to generate novel street tags that transcend traditional letter forms. The writer is able to live-generate her own ML graffiti tags in the ghostly belly of a deep convolution neural network and then transfer them to a real surface using spray paint. The result is an interactive installation of ML augmented tags in physical space.

02. Machine Learning and Graffiti
"The "Amazing Augmented Tagger Machine" is a hybrid of a custom-made piece of software and an interactive installation, based on a deep convolution neural network, trained on a data set of ~1,000 annotated street tag images from Germany. The model is able to generate new graffiti tags based on any textual input from the audience.

A sample of the training data set

03. Initial Results

Left column: annotation, Center column: generated tag, Right column: original tag

04. Style Augmentation
The generated output by the machine learning model can subsequently be used to transfer tags back onto a physical surface using spray paint, serving as a source of inspiration for new letter forms, connections and styles. An initial test produced a real-world collage in the shape of a canvas of different tags produced by the writer in collaboration with the machine.