FreshU-GAN: A Deep Learning model for determining beef freshness

Meat freshness is an important aspect of food safety. As one of the most common types of meat, accurate assessment of beef freshness helps protect consumers’ health and prevent potential health risks. To provide a convenient and accessible method for consumers to evaluate beef freshness based solely on visual information, we propose a novel deep learning framework that creatively integrates U-Net and Generative Adversarial Networks (GANs). Specifically, U-Net serves a dual purpose: as the generator within the GAN to produce realistic samples, and as a feature extractor for freshness classification. The discriminator in the GANs compels the U-Net to learn meaningful and discriminative features that improve classification performance. To validate the robustness and adaptability of our model, we executed our model on three different individual datasets, as well as the pooled dataset, to demonstrate the effectiveness and versatility of our proposed model across various imaging conditions.