Florian van Zandwijk

Florian van Zandwijk (he/him)

Vision Processor: EN471 (2019)

Vision Processor: EN471 (2019), video installation, Vision Processor chip, machine learning, video dataset, found footage, Rotterdam, The Netherlands

Alt text: Installation in an exhibition space. Made from black scaffolding tubes at different heights with a screen on each support. The three smaller outside screens show video’s of archetypal physical behavior for protestors, construction workers and traffic controllers wearing yellow vests. They are surrounded by grass. While the center screen shows existing video footage of protestors and the classification percentages from the machine learning algorithm.

Construction Worker

Alt text: Compilation of archetypal moves for construction workers performed by multiple people in a grass field, all wearing a white t-shirt, blue jeans and a yellow vest. Some of the performers wear typical construction gloves, hats and shoes.

Protestors

Alt text: Compilation of archetypal moves for yellow vest protestors in a grass field, performed by multiple people in a yellow vest, wearing mostly black pants and tops, some of the protestors covering their faces with masks or black t-shirts. The video goes from full screen to a collage of multiple people doing the same gestures.

Traffic Controller

Alt text: Compilation of archetypal moves for traffic controllers performed by multiple people in a yellow vest, wearing high visibility caps and standing in a grass field. The video goes from 1 person full screen to a group setting, to a collage of multiple people doing the exact same typcial gestures.

vision processor

Alt text: A square computer chip in the center of the image attached to black foam background.

Classification

Scrolling timeline of real footage from protestors, traffic controllers, construction workers and everything in between. The small video’s are placed left of the screen and the algorithm classifies each video with certain percentages for the protestors, traffic controllers, construction workers. This is shown next to the video’s through yellow bars with percentages on a black background. In the right top corner a text reads Vision Processor in white with EN471 in yellow.

Since the emergence of the yellow vests movement in France on November 17, 2018, the yellow vest has rapidly transformed from a signifier of authority and safety to the symbol of a grassroots movement for economic justice. Vision Processor: EN471 uses a video-dataset which illustrates three archetypal physical behaviours of people wearing yellow vests. A machine learning algorithm trained with this dataset tries to classify existing footage of these roles within their existing context and meaning, which can easily be understood by humans. Although computer vision is easily able to classify a human being, will it ever be able to classify its intentions? This work asks whether a computer vision algorithm will ever be capable of distilling context out of data.