aiqbench analysis
This research proposes a novel method for generating geometric scene graphs to represent assembly relationships between components using minimal data and no semantic priors. By leveraging Faster R-CNN outputs with transformer architectures and attentional graph convolution networks, the approach aims to enhance robotic understanding of how parts fit together in manufacturing contexts. While framed around consumer product assembly, the core technical contribution lies in advancing computer vision techniques for structural reasoning within robotics systems.
Geometric 2D Scene Graph Generation
arXiv:2607.22325v1 Announce Type: cross Abstract: In production processes for consumer products, assembly instructions are essential not only for planning but also for executing the production process. Likewise in robotics, it is crucial for an assembly robot to understand how components fit together and can be assembled. To facilitate these tasks, we contribute a method for constructing scene graphs to represent and characterize assembly relationships between components. Our approach does not rely on semantic data and is capable of handling a very small dataset. To realize this, the output of a Faster R-CNN model is used to create geometric representations, which are then processed by a transformer architecture to generate an adjacency matrix. This matrix serves as input to a Siamese network that uses message passing based on an attentional graph convolutional network (aGCN) architecture to characterize the connections between the components. We validate our method on a study dataset of toy model components which can be assembled into transportation vehicles.
Abstract:In production processes for consumer products, assembly instructions are essential not only for planning but also for executing the production process. Likewise in robotics, it is crucial for an assembly robot to understand how components fit together and can be assembled. To facilitate these tasks, we contribute a method for constructing scene graphs to represent and characterize assembly relationships between components. Our approach does not rely on semantic data and is capable of handling a very small dataset. To realize this, the output of a Faster R-CNN model is used to create geometric representations, which are then processed by a transformer architecture to generate an adjacency matrix. This matrix serves as input to a Siamese network that uses message passing based on an attentional graph convolutional network (aGCN) architecture to characterize the connections between the components. We validate our method on a study dataset of toy model components which can be assembled into transportation vehicles.
Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
I.2.10; I.4.8; I.5.1
arXiv:2607.22325 [cs.CV]
(or arXiv:2607.22325v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2607.22325 arXiv-issued DOI via DataCite (pending registration)
Jahn, C., Waldmann, U. and Goldluecke, B. (2026). Geometric 2D Scene Graph Generation. In Proceedings of the 18th International Conference on Agents and Artificial Intelligence - ICAART 2026
https://doi.org/10.5220/0014249800004052 DOI(s) linking to related resources
Full text under the publisher's open licence · 215 words
Related coverage
SG-Layout: Structured Scene Graph-Guided Layout Generation with LLMs
arXiv · Aug 5, 2026
Generation of High-Level Concepts in 3D Scene Graphs via Autoregressive Diffusion
arXiv · Sep 2, 2026
VIGOR: Verifiable Industrial Geometry Optimization and Repair in Scene Programs
arXiv · Oct 12, 2026
From Scene-Centric to Observer-Centric: Modeling Observer-Aware Relations for 3D Scene Graph Generation
arXiv · Jul 21, 2026