From glow to insight: user-independent AI and edge computing for ultra-fast, accurate contamination detection in food facilities

Résumé fourni par la source

Handheld fluorescence tools are widely used for hygiene inspection but often suffer from operator subjectivity, lighting variability, and false positives. This work presents the SafetySpect CSI platform, which uses onboard edge computing and AI to deliver consistent, objective contamination detection. Three large, QA-validated UV fluorescence datasets were collected across kitchen, food processing, and meat processing environments to train and test segmentation models. Among YOLO-based and transformer-based approaches, SegFormer achieved the strongest performance (mIoU 0.83–0.85) with high-speed inference. Results show that combining curated datasets with advanced AI enables fast, accurate, and generalizable contamination detection suitable for next-generation industrial fluorescence imaging systems.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.