Enhancing fruit freshness classification with adaptive knowledge distillation and global response normalization in convolutional networks

The assessment of fruit freshness is crucial for ensuring food quality and reducing waste in agricultural production. In this study, we propose <i>Global Response Normalization and Gaussian Error Linear Unit Enhanced Network (GGENet)</i>, a novel deep learning architecture that leverages adaptive knowledge distillation (AKD) and global response normalization (GRN) to classify fruits as fresh or rotten. Our model comprises two variants: <i>GGENet-Teacher (GGENet-T)</i>, serving as the teacher model, and <i>GGENet-Student (GGENet-S)</i>, functioning as the student model. By transferring attention maps from the teacher to the student model, we achieve efficient adaptive knowledge distillation, enhancing the performance of the lighter student model. Experimental results demonstrate that the <i>GGENet with adaptive knowledge distillation (GGENet-AKD)</i> achieves a competitive accuracy of 0.9818, an F1-score of 0.9818, and an area under the curve (AUC) score of 0.9891. The proposed method significantly contributes to reducing food waste and enhancing quality control in agriculture by facilitating early detection of rotting fruits.