Reproduce “GBOT: Graph-Based 3D Object Tracking for Augmented Reality-Assisted Assembly Guidance” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shiyu Li, Hannah Schieber, Niklas Corell, Bernhard Egger, Julian Kreimeier, Daniel Roth
Researchers from Technical University of Munich and Friedrich-Alexander-Universität Erlangen-Nürnberg developed GBOT, a graph-based 3D object tracking system designed for Augmented Reality assembly guidance, which achieves 92.4% ADD/ADD-S accuracy with a 2.6 cm translation error in real-time, even under challenging conditions. The work also introduces the GBOT dataset, a synthetic RGB-D benchmark for dynamic multi-object assembly scenarios.
Key insights
- Integrating initial 6D pose estimation (via adapted YOLOv8Pose) with continuous graph-based tracking effectively addresses both real-time performance and robustness for AR assembly.
- The proposed multi-state assembly graph, which dynamically adapts to changing kinematic relationships, improves tracking accuracy and resilience to occlusions during complex assembly sequences.
- Generating a comprehensive synthetic dataset with extensive domain randomization, including hand occlusions and varied lighting, is essential for training models that perform well in real-world AR environments.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.