simittag

what

Simittag is a marker system for robots and cameras that need to know exactly where they are.

Point a camera at a tag and you get its full position and orientation from a single frame, along with whatever the tag carries: an id, a short piece of text, even a GPS position. Tags work black on white or white on black, and both can show up in the same frame.

variants

Multiple variants for a plethora of use cases.

variantgridpayloadids
s4k3×1612-bit id4,096
s2563×161 byte256
s64k3×1616-bit id65,536
s16m4×244 bytes16.7 M
sdata5×3611 bytes2⁸⁸

For most projects we'd start with s16m. If the tag itself needs to hold text, a namespace or a location, sdata fits up to 11 bytes.

pose

Get pose with industry-leading accuracy.

0.03°tilt error
0.07°rotation error
0.04 %depth error
0.6 pxcenter error

* tests conducted with an s16m tag tilted from 0 to 70 degrees

range

Printed on a sheet of A4, a tag reads from more than 8 meters away.

system1280 px1920 px
simittag s4k8.4 m12.6 m
simittag s16m6.7 m10.0 m
simittag sdata5.6 m8.5 m
AprilTag 36h119.6 m14.4 m
ArUco 6x610.0 m15.0 m

Measured with a 175 mm tag and a 60° camera, counting only distances where at least 90% of frames decode.

Simittag decodes with up to 30% coverage, where competitors fail after 5%.

speed

Fast enough to run on every frame.

5 msrust
23 mswebassembly
65 mspython

Six tags in a 1280×1280 frame on a 12-core ARM processor.

install

There's a Python reference implementation, a dependency-free Rust detector that also builds to WebAssembly, and a ROS 2 node. Every release comes with prebuilt binaries for Linux, macOS and Windows.

license

Simittag is BSD 2-Clause licensed, and the tag format is free for anyone to implement.