This website allows you to take a selfie, which will then be analyzed using Artificial Intelligence (AI). The AI system will classify the image either as a human or as a certain object or animal. The goal is to trick the AI, so that you are not identified as a human being. You can, e.g., use your body, clothes or make-up to fool the AI. The site is designed for education, research and entertainment.

The AI components used for the image analysis consist of a PoseNet model (that estimates the pose of a human in an image) and a MobileNet image classifier. Both have been implemented using the wonderful ml5.js library, which builds on top of TensorFlow.js. The rest of this site has been built with standard web technologies and the JavaScript library for creative coding p5.js.

The image processing happens locally, on your computer and in your browser. Video images are not transmitted to us or to third parties.

If you have questions, suggestions or concerns, please contact Hanna Schraffenberger.