The Interpretive Gallery we designed and developed for SFMOMA launched as part of the 2018 Magritte: The Fifth Season exhibition in 2018. After the launch we learned that the marketing department was counting visitor images shared on Instagram and other social networking sites by hand in order to understand how visitors were experiencing and sharing the space. We thought this process could be automated and build a POC using Tensorflow to count images shared at certain hashtags. We would download images shared at hash tags like #magritte and #sfmoma and were able to sort which ones were taken in the interpretive gallery with about 70% accuracy which would sort them from images shared at the hashtags from thousands of others taken around the world. We were also able to train the model to understand which of the six experiences in the space the image was take that would tell us how often the experiences were shared relative to one another.
