ArUco Nano: a simpler, faster, and more reliable fiducial marker detector

Squared fiducial markers are essential for many augmented reality and robotics applications, with ArUco serving as the de facto standard, largely due to its integration with OpenCV. However, the standard implementation relies on general-purpose algorithms that are not specifically tailored to this task, resulting in significant computational bottlenecks, especially on resource-constrained devices. We introduce ArUco Nano, a minimalist (500 lines of code), header-only C + + library that achieves a speedup of up to 6.5 × over the standard OpenCV library. By implementing a novel Visited-Aware contour extraction algorithm and direct sub-pixel code sampling, the library avoids computationally expensive contour extraction and image warping processes while also improving detection reliability. Benchmarks on up to 16 MP imagery demonstrate that ArUco Nano consistently outperforms the current OpenCV implementation in our evaluation, achieving higher speed and a higher F1 score.