SIMULACRAMOMA

Maximum object detection on the Moma collection.

January 2017 - online

SIMULACRAMOMA is a playful look at object recognition systems, asking them to dive deep into modern art paintings from the MOMA collection. Usually throttled to a respectable degree of certainty, the algorithm has been set free to interpret each paint stroke or shape it thinks it recognises. Sadly my poor robot has never seen such abstract art before, having trained on a more representational data set, and while it may find meaning in these artefacts of modernity, it can only be articulated in the blandest terms.

The work comprises a selection of famous paintings from the MOMA collection which have been passed through the Darknet YOLO object detection system with a excessively low tolerance of 0.001. The output is a render of the painting with hundreds of objects recognised, very few of which a human would notice, marked with the traditional crude bounding boxes and labels. The original painting has been defaced to the point of incoherence with occasional features popping out.

The work lives as a Flickr set of 517 pieces. Here are five.