<![CDATA[The advancement of artificial intelligence and computer vision has enabled automated quality assessment in agricultural products to reduce subjectivity and inefficiency in manual inspection. This study aims to compare the performance of two transfer learning architectures, MobileNetV2 and Xception, for fruit and vegetable freshness classification. The dataset was obtained from Kaggle and focused on eight classes: fresh and rotten apples, bananas, cucumbers, and tomatoes. To address data imbalance and improve model generalization, data augmentation techniques including rotation, horizontal flipping, and vertical flipping were applied. Both pre-trained models were enhanced by adding a Global Average Pooling layer, a 128-neuron Dense layer, and Batch Normalization. The models were trained using the Adam optimizer with a learning rate of 0.0001 for 20 epochs. Performance evaluation was conducted using accuracy, precision, and recall metrics. Experimental results show that MobileNetV2 achieved an accuracy of 99.43%, while Xception obtained 99.39%, with consistently high precision and recall values. These findings demonstrate that the transfer learning approach is highly effective for fruit and vegetable freshness classification and provides competitive performance compared to previous studies.]]>
Copyrights © 2026
