<![CDATA[Accurate classification of beef and pork images is crucial for developing reliable automated food inspection systems, particularly due to their visual similarity in color, texture, and muscle fiber patterns. This study aims to comparatively evaluate the performance of multiple deep learning models for binary meat image classification using RGB digital images. Four Convolutional Neural Network (CNN) architectures, namelyInceptionV3, VGG16, ResNet50, and Xception were assessed under identical preprocessing pipelines and hyperparameter settings to ensure a fair comparison. The dataset underwent cropping, resizing to 224×224 pixels, normalization, and augmentation to enhance variability and improve generalization performance. Model effectiveness was measured using accuracy, precision, recall, and F1-score on unseen test data. Experimental results show that InceptionV3 achieved the most balanced classification performance, with a test accuracy of 72% and an F1-score of 0.7. Although Xception obtained higher training accuracy, it exhibited overfitting during testing, while VGG16 and ResNet50 demonstrated comparatively lower classification capability. These findings indicate that InceptionV3 provides a more stable and generalizable architecture for beef and pork image classification. The study highlights the importance of cross-architecture evaluation in developing robust CNN-based systems for automated meat classification.]]>
Copyrights © 2026
