Motivation:

The progress of multimodal foundation models continues to be impressive. Vision-Language Models (VLMs) can not only process text but also interpret and understand visual context from images. As a result, they offer a novel approach for visual industrial anomaly detection tasks.

Goal of the thesis:

The goal of this thesis is to develop a mobile application for Automated Optical Inspection (AOI) that utilizes a Vision-Language Model (VLM) in the backend.

Work Packages:

  • Literature Review
    • Research on the utilization of VLMs in industrial anomaly detection tasks
  • Market Research
    • Investigation of mobile app usage in manufacturing, with a focus on quality control and Automated Optical Inspection (AOI).
  • App Development
    • Development of a prototype mobile app (Android or iOS).
  • Model Testing
    • Evaluation of multiple VLMs using a one-shot prompting approach (e.g., OpenAI, Gemini, etc.).
  • Performance Evaluation
    • Assessment of AOI performance using benchmark datasets such as MvTec.
  • Hosting and Inference Evaluation
    • Comparison of different hosting options
      • On-device (local on the smartphone)
      • Internal FAPS server
      • Cloud providers (e.g., AWS)

Your Profile:

  • Experience in coding (app development and GenAI experience are strong advantages)
  • Independent and structured working style
  • Ability to document your work clearly and thoroughly
  • Able to attend weekly scrum meetings onsite at FAPS Nuremberg (mandatory)

Required Application Documents:

  • Curriculum Vitae (CV)
  • Transcript of Records
  • Gantt Chart outlining your proposed project timeline
  • One-pager describing your motivation and reasoning on why this approach (VLM and mobile app for AOI) could be beneficial
    • One-pager generated purely by ChatGPT will be ignored. 

In case you are invited to an online meeting, please additionally prepare by reading this article

Incomplete or generic applications will not be considered.

Kategorien:

Forschungsbereich:

Signal- und Leistungsvernetzung

Art der Arbeit:

Bachelorarbeit, Masterarbeit, Projektarbeit

Kontakt:

Bernd Hofmann, M.Sc.

Koordinator Technologiefeld Qualität und Management

Department Maschinenbau (MB)
Lehrstuhl für Fertigungsautomatisierung und Produktionssystematik (FAPS, Prof. Franke)