May 2017
Dotframes
by Josh Horowitz, Toby Schachman
This entry is a stub, and will be expanded on a future archiving pass.
The first and primary fiducial used in Realtalk for object identification and tracking.
Each corner of the object has a wedge of five dots. The color pattern encodes the page number, in a way that is robust to changing lighting conditions.
The recognizer is simple and easy to understand. Allows for objects of widely varying sizes on the same surface, even overlapping. Fairly robust to occlusion.
Invites improvisation! Resize a page by ripping it in half, or rip the corners off and put them around a posterboard or table, or make your own dotframe with colored markers or stones, etc.
Failures are understandable. When it's not working, it's because a corner is covered up, or creased, or there's inadequate lighting, etc. The problem can be discovered and corrected in the physical world.
Earlier: Framecodes (2016)
Later: Tiny dotframes (2022)
Earliest making-of:
