ARTag

Overview

ARTag is a fiducial marker system for 3D registration (alignment) and pose tracking in augmented reality [3]. It is not a data-carrying barcode: each marker encodes only a small identifier — at most one of 2002 distinct IDs — and the system's purpose is to find that marker robustly in an image and recover the camera's pose relative to it [3]. A marker is a bi-tonal (black-and-white) planar pattern: a square outline enclosing a 36-bit digital word in the interior [1]. Like apriltag and aruco, ARTag trades data capacity for robust, quantifiable detection.

The system was developed by Mark Fiala at Canada's National Research Council (NRC) [1, 3].

History

ARTag was introduced by Mark Fiala around 2004–2005. The NRC technical report "ARTag Revision 1. A Fiducial Marker System Using Digital Techniques" (report number NRCC 47419) is dated November 2004 [1], and the peer-reviewed paper "ARTag, a fiducial marker system using digital techniques" was published at IEEE CVPR 2005 [2]. ARTag was explicitly positioned as an improvement on the template-matching markers of artoolkit-marker, replacing stored reference images with a digitally coded ID protected by error-correcting codes [3].

Technical specification

Marker structure. An ARTag marker is a "bi-tonal planar pattern that consists of a square outline with a digital 36-bit word encoded in the interior" [1] — i.e. a 6×6 grid of black/white cells inside a black border.

Coding and error correction. Rather than storing patterns, ARTag assigns each marker a unique ID protected by digital coding theory using "checksums and forward error correction (FEC)" [1]. This gives a low, numerically quantifiable false-positive and inter-marker-confusion rate [3]. The marker library holds 2002 unique IDs [3].

Detection. ARTag uses an edge-linking method to find marker boundaries, which gives robustness to lighting variation and partial occlusion and removes the need for a global grayscale threshold that earlier systems required [1, 3]. Because IDs are coded, no per-marker training is needed.

Pose recovery. Once the square's four corners are located and the ID decoded, the marker corners feed standard pose estimation to recover the camera's position and orientation relative to the marker, allowing virtual content to be registered at the marker's location [3].

Use cases

• Augmented reality — registering virtual objects, games and animations onto physical markers [3].
• Robot navigation and SLAM — fixed visual references of known pose [3].
• Space outreach / engineering — see Fun facts.

Comparison

Versus artoolkit-marker: ARTag replaces ARToolKit's template matching with a digitally coded 36-bit ID, eliminating the grayscale threshold, lowering false positives, allowing smaller markers, and adding occlusion robustness [3]. Versus aruco and apriltag: ARTag is the conceptual predecessor of these square-binary, error-corrected fiducials — all share the black-bordered grid with a coded ID and edge-based quad detection — but ARTag's library is comparatively small (2002 IDs) and the reference implementation is historic/proprietary rather than the open, actively-maintained libraries that ArUco and AprilTag became [3].

Fun facts

• An ARTag appears on NASA's Mars Science Laboratory (Curiosity rover) [3].
• NASA used a similar technique in its Spacecraft 3D smartphone app as an educational outreach tool [3].
• ARTag is supported by the open-source Goblin XNA software framework [3].

Status

Historic / largely superseded. ARTag was influential — it established the digitally coded square fiducial that aruco and apriltag later popularized — but it is no longer the active fiducial of choice; ArUco and AprilTag dominate current robotics and AR use [3].

Sources

1. ARTag Revision 1, A Fiducial Marker System Using Digital Techniques (NRCC 47419) — NRC Publications Archive (Canada), 2004
2. ARTag, a fiducial marker system using digital techniques — IEEE CVPR, 2005
3. ARTag — Wikipedia, accessed 2026

Deployments

No country reports mention this standard by name.

Regions / aggregations not mapped to a single country

• Universal