Computer Vision | Additive Manufacturing | Human-Computer Interaction
Invisible Markers
During an internship at the Matter of Tech Lab at Cornell Tech, this project focused on bridging the gap between hardware assembly and user accessibility.
That Part There was developed to simplify the process of identifying and debugging hardware components for non-experts. By embedding fiducial markers into 3D-printed assembly parts, the system allows users to point to a loose or broken component, enabling easy identification and remote troubleshooting without requiring detailed technical knowledge.
2023
I evaluated multiple kinds of fiducial markers that would allow That Part There to be more reliable and easy to use, as well as identifying the best markers to 3D print.
The project involved testing fiducial markers to ensure they were both effective and could be integrated into the hardware components. Various marker designs were created using CAD tools and 3D-printed with multi-material techniques, incorporating both regular black PLA and infrared-sensitive PLA for optimal detection. These markers were then embedded into assembly components, allowing them to remain visually unobtrusive while still highly functional for computer vision systems.
The black tiles contain the markers that are ‘invisible’ to the eye. However an infra-red camera can easily detect them. The evaluated markers include: AprilTag | STag | reacTIVision | arUco
Testing robustness
What Does ‘Robustness’ Mean?
In the context of this project, robustness refers to the ability of the system to remain reliable even when exposed to disturbances. For example, in the video, the marker is deliberately disturbed by a finger, simulating real-world interference. The video demonstrates the performance of an STag, one of the most resilient markers tested. Unlike others, the STag maintained recognition even when its edges were interrupted, showcasing its superior robustness under challenging conditions.
To evaluate the detection range of the markers, we tested how far a marker could be recognized by the system. In the videos, various types of 3D-printed April Tags are shown, while an infrared camera is raised and lowered to determine the maximum and minimum range of detection. This process helped identify the optimal markers and configurations for reliable performance across varying distances, ensuring the system's effectiveness in real-world applications.
Testing Detection Range
Testing Saturation
Saturation testing involved determining how many markers could be embedded within a single component without compromising detection. This process assessed both the minimum size of the markers and their ability to be simultaneously recognized when placed in close proximity. The goal was to optimize marker density while ensuring reliable detection, paving the way for efficient integration into complex hardware assemblies.
CREDITS
Maia Hirsch
PhD. Amrit Kwatra
Prof. Thijs Roumen
Matter of Tech Lab,
Cornell Tech
Software + 3D printing
Camera Assembly
Advisor
Facilities