Team: Shunya Kato (Research Assistant), Yuichiro Takeuchi
We are developing a new class of 2D visual codes called Ninja Codes, intended to be used for applications such as augmented reality and robotics. By harnessing the power of adversarial example generation — a technique borne out of deep learning research that produces images that appear to show entirely different content to humans and computers — Ninja Codes are rendered discreet, concealed to human eyes but easily recognizable to detection software. The codes can be used to realize precise location tracking in environments where placing (often conspicuous) conventional visual codes is undesirable.
Publications:
- Takeuchi, Y. Ninja Codes: Exploring Neural Generation of Discreet Visual Codes. In Ext. Abst. of CHI 2021. No. 224. Download (PDF, 1.8MB)