EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Object Centric Visual Question Answering for Robotic Assembly Inspection and Workflow Fault Diagnosis Tasks

Marina Alvarez, Ines Pereira, Hannah Becker

Journal of Clinical Practice and Health Sciences · 20.06.2026 · 6 (1) · 131-141

DOI: 10.71465/jcphs53 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Clinical Practice and Health Sciences, online erschienen am 20.06.2026. Verfasst von Marina Alvarez und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The rapid evolution of modern manufacturing paradigms necessitates highly adaptable and intelligent robotic assembly inspection cells. Traditional machine vision systems rely on rigid, pre-programmed rules and spatial templates, which frequently fail when confronted with unstructured environments, diverse product portfolios, or unexpected anomalies. This paper proposes a novel framework centered on Object-Centric Visual Question Answering to address the growing demand for flexible and cognitive inspection methodologies. By treating the robotic cell as a dynamic environment where human operators can query the system using natural language, the proposed framework enables intuitive interaction and diagnostic transparency. Unlike standard global-feature visual question answering models, our approach parses visual scenes into discrete, object-centric representations. This allows the system to ground natural language queries to specific mechanical components, fasteners, and spatial relationships within the assembly workspace. The research details the architectural design, object discovery mechanisms, and cross-modal reasoning required to achieve high-fidelity inspection results. Extensive evaluations on a custom dataset of robotic assembly scenarios demonstrate that the object-centric approach significantly outperforms global feature baselines in terms of accuracy, interpretability, and robustness against occlusions. Ultimately, this framework paves the way for more collaborative, human-centric robotic manufacturing systems capable of complex reasoning and real-time fault diagnosis.

Zitieren

APA

Alvarez, M., Pereira, I., & Becker, H. (2026). Object Centric Visual Question Answering for Robotic Assembly Inspection and Workflow Fault Diagnosis Tasks. Journal of Clinical Practice and Health Sciences, 6(1), 131-141. https://doi.org/10.71465/jcphs53

Vancouver

Alvarez M, Pereira I, Becker H. Object Centric Visual Question Answering for Robotic Assembly Inspection and Workflow Fault Diagnosis Tasks. Journal of Clinical Practice and Health Sciences. 2026;6(1):131-141. doi:10.71465/jcphs53

BibTeX
@article{Alvarez2026,
  author = {Alvarez, Marina and Pereira, Ines and Becker, Hannah},
  title = {Object Centric Visual Question Answering for Robotic Assembly Inspection and Workflow Fault Diagnosis Tasks},
  journal = {Journal of Clinical Practice and Health Sciences},
  year = {2026},
  volume = {6},
  number = {1},
  pages = {131--141},
  doi = {10.71465/jcphs53}
}
Kostenlos starten →