A web-based platform for real-time stewed beef freshness monitoring: Integrating anthocyanin colorimetric film with deep learning.

This study developed a web-based online monitoring system for assessing the freshness of stewed beef by integrating an intelligent colorimetric film with deep learning. A pH-responsive colorimetric film was fabricated using mulberry anthocyanin extract (MAE) as the indicator, gelatin (G) and carboxylated cellulose nanofibers (CCN) as the film-forming matrices, and Mg2+ as the cross-linker. When applied to monitor stewed beef stored at 4 °C, the film exhibited distinct color changes correlated with spoilage indicators. An improved Ordinal-ResNet-50 model, incorporating a custom ordinal loss function, was trained on the film images to classify freshness into three grades ("fresh", "less-fresh", and "spoiled"), achieving 98.36% accuracy across three freshness grades. Based on this model, a user-friendly web-based monitoring system was implemented, enabling real-time and non-destructive freshness assessment within 1 s after film image upload. This work establishes a complete technical framework, from smart sensing to online intelligence, for digital food safety management.