Focus on SMEs experiments: Quality Assurance of clothing production

WHAT ABOUT THE EXPERIMENT? 

​

As part of AI REDGIO 5.0, Katty Fashion is leading this experiment to transform traditional quality control practices through AI-powered automation. In a field where inspection is often manual and subjective, this initiative aims to bring consistency, speed, and intelligence to the process by introducing edge-based AI systems capable of detecting clothing defects in real time. 

The system captures product images on the shop floor and uses fuzzy logic to differentiate between defective and non-defective items. The goal is to streamline quality checks while building a comprehensive defect database that helps improve production workflows over time. 

During the first iteration, Katty Fashion successfully deployed its edge infrastructure and began collecting data onsite. In partnership with the Gheorghe Asachi Technical University, the team developed the fuzzy AI logic needed for handling complex garments. The experience also revealed important lessons: challenges around data quality, model training efficiency, and repeatability highlighted the need for a stronger technical foundation in future phases. 

 

WHAT IS EXPECTED? 

​

The second iteration of the experiment aims to deliver a fully integrated, tested, and validated TRL 7 prototype that becomes part of Katty Fashion’s production workflow. This phase includes augmenting the fuzzy AI with a convolutional neural network to handle more complex scenarios and improve defect classification accuracy. 

The team is also working on creating a dedicated infrastructure for model training and retraining, along with a streamlined data pipeline for preparing high-quality datasets. By combining edge computing, AI, and real-time monitoring, the solution is set to enhance quality assurance while supporting Katty Fashion’s long-term vision of becoming a digitally enabled, efficient, and circular fashion manufacturer.