FidX

FidX is a cross-platform fiducial marker library and runtime designed for detection, pose estimation, and mapping in robotics and augmented reality. It provides a modular pipeline, including marker design, image processing, pose estimation, and an optional mapping layer. The project emphasizes speed, accuracy, and robustness to occlusion and lighting variation, and offers bindings for C++, Python, and other languages.

FidX originated in the Vision Lab at Tech University in 2014 and was released as an open-source project in 2016. It has since undergone several major revisions, incorporating improved marker families and calibration tools, with ongoing contributions from universities and robotics companies. It is widely used in academic courses and pilot systems.

Architecture: The core components include a detector module, a pose estimator, a marker database, and a mapping layer. Detectors operate on grayscale images and return 3D camera poses with confidence scores. The core is written in C++ with bindings for Python and other languages, and it integrates with ROS and OpenCV.

Marker families and robustness: FidX supports multiple marker families designed to maximize distinguishability and error tolerance, including FidX-Quad and FidX-Grid. Each marker carries a unique ID and error-correcting code to improve recognition under partial occlusion or blur. The library supports multi-marker pose estimation and drift-free map updates.

Applications: The system is used for robotics navigation and mapping, augmented reality experiences, drone docking, and industrial automation. It is also employed in education and research to study localization, SLAM, and computer vision algorithms.

Licensing and community: FidX is released under an open-source license compatible with academic and commercial use. It maintains a public code repository, issue trackers, and contributor guidelines, with active documentation and example projects to aid adoption.