Deep learning-assisted portable fluorescent sensing array based on multiple carbon dots for rapid seafood freshness assessment.

Given the highly perishable nature of seafood, real-time monitoring its freshness is crucial for ensuring food safety and minimizing waste. However, current fluorescent sensing array (FSA) systems are often hindered by aggregation-caused quenching (ACQ), insufficient recognition accuracy, and limited platform applicability. Herein, the FSA based on aggregation-induced emission (AIE)-active carbon dots (CDs) was developed for monitoring total volatile basic nitrogen (TVB-N) markers of seafood freshness. The cross-reactive process between nine AIE-CDs and ammonia endowed the FSA with excellent detection performance (limit of detection of 5 ppm). By integrating the FSA with a deep convolutional neural network (DCNN), the trained lightweight DCNN model MobileNetV3-Large achieved an overall accuracy of 97.54% for freshness classification. For practical implementation of MobileNetV3-Large, a smartphone application with a user-friendly interface was developed for image acquisition and result visualization, providing an integrated, end-to-end platform for the intelligent and rapid monitoring of seafood freshness.