Quote of the day

A person who never made a mistake never tried anything new.

- Albert Einstein

Contact

Few-Shot Vision-Language Inspection with Verifiable Explanations on Production Lines

Open Industry & Manufacturing

Few-Shot Vision-Language Inspection with Verifiable Explanations on Production Lines

Few-Shot Vision-Language Inspection with Verifiable Explanations on Production Lines

Description

Detect surface defects from very few labeled examples using vision-language models while ensuring that generated explanations are spatially consistent with the actual defective region.

Details

Context and Problem Statement

Industrial quality inspection suffers from extreme defect scarcity and frequent product changes. Vision-language models can support few-shot inspection, but their textual explanations may not correspond to the actual defect location, reducing operational trust.

Research Question

How can consistency among classification decision, defect localization, and textual explanation be aligned and verified in a few-shot vision-language inspection system?

Proposed Approach

Adapt vision-language models through prompt learning and normal-only examples, add a localization head, verify agreement between anomaly maps and generated descriptions, and distill the system into a compact model for real-time deployment.

Expected Contribution

A quantitative protocol for evaluating explanation faithfulness in industrial anomaly detection and the performance cost of enforcing consistency.

Expected Prototype

A real-time inspection station that highlights suspicious regions, produces corresponding explanations and confidence scores, and supports rapid addition of new defect types.

Datasets

MVTec AD, VisA, and partner-line images annotated by experts.

Challenges

Extreme class imbalance, lighting variability, latency constraints, expert-annotation cost, and plausible but unfaithful explanations.

Research Question

Can a few-shot vision-language inspection system guarantee faithful agreement between textual explanation and defect localization?

Innovation

The project treats explanation as a verifiable output rather than a cosmetic interface feature.

Expected Deliverable

A real-time inspection prototype, an explanation-faithfulness evaluation protocol, and an annotated industrial image dataset.

Technologies

Vision-Language Models (CLIP, LVLM) Zero-Shot and Few-Shot Anomaly Detection Knowledge Distillation and Real-Time Inference Explainable AI and Localization Maps Industrial Vision

Required Skills

  • Computer Vision and Multimodal Models
  • Real-Time Optimization and Deployment
  • Industrial Vision and Lighting
  • Experimental Evaluation Methodology

Datasets

  • MVTec AD
  • VisA (Visual Anomaly dataset)
  • Images collected from a partner production line

Morocco & Africa Relevance

Moroccan manufacturing sites frequently change product references; few-shot inspection can reduce the annotation burden and deployment cost of industrial machine vision.