Vision–Language AI for Automated Manufacturing Quality Inspection

#Vision–Language #Models #Fault #Detection #AI #LLMs #Automated #Manufactoring
Share

Is AI ready to identify a faulty component, pinpoint the problem, and run on a compact device such as a smart camera without human-intervention?

This talk presents two studies of industrial quality inspection. The first compares AI methods that rely on visual information with vision–language approaches that combine images and written descriptions, examining accuracy, reliability, fault localization, and computational requirements. The second explores how written descriptions, which can be generated using familiar AI tools such as ChatGPT and Perplexity AI, can guide the identification of faulty components without examples of faulty products or retraining the underlying model. Together, these studies examine AI’s inspection capabilities and the practical challenges of deployment on devices with limited computing resources.



  Date and Time

  Location

  Hosts

  Registration



  • Add_To_Calendar_icon Add Event to Calendar
  • 401 Sunset Avenue
  • University of Windsor
  • Windsor, Ontario
  • Canada
  • Building: Erie Hall
  • Room Number: 1118

  • Contact Event Host
  • Starts 06 November 2026 03:00 PM UTC
  • Ends 20 November 2026 06:30 PM UTC
  • No Admission Charge


  Speakers

Atefeh Gilvari

Topic:

Vision–Language AI for Automated Manufacturing Quality Inspection

Is AI ready to identify a faulty component, pinpoint the problem, and run on a compact device such as a smart camera without human-intervention?

This talk presents two studies of industrial quality inspection. The first compares AI methods that rely on visual information with vision–language approaches that combine images and written descriptions, examining accuracy, reliability, fault localization, and computational requirements. The second explores how written descriptions, which can be generated using familiar AI tools such as ChatGPT and Perplexity AI, can guide the identification of faulty components without examples of faulty products or retraining the underlying model. Together, these studies examine AI’s inspection capabilities and the practical challenges of deployment on devices with limited computing resources.

Biography:

Atefeh is a PhD candidate working on camera-based defect detection for manufactured components, with a focus on electric vehicle parts where subtle defects may indicate early safety risks. Her research develops fast, data-efficient methods that run directly on industrial cameras, enabling reliable real-time inspection with minimal training and computation.