Integrating Convolutional Neural Network for Freshness Evaluation of Pork

When purchasing pork, traditional methods such as odor assessment and visual inspection are the most commonly employed approaches by consumers. However, these methods are highly subjective, often unreliable, and fail to guarantee the hygiene and safety of the pork. Additionally, they require significant expertise to effectively determine the quality and freshness of the meat. To address these challenges, this study aims to develop an automated and objective method for evaluating pork freshness. The proposed approach integrates two advanced convolutional neural networks (CNNs), namely ResNet-50 and Inception_v3, to enhance the feature extraction process for pork freshness assessment. Both networks are modified using transfer learning techniques to better capture critical features from the input images. The extracted features from these two networks are fused, enabling the complementary strengths of the individual networks to be fully utilized. A SoftMax classifier is employed to classify the fused features into freshness categories, with the entire model being trained using the cross-entropy loss function and the gradient descent optimization algorithm. The integration of these methods and techniques results in a highly efficient and accurate model. Experimental results demonstrate that the proposed model achieves a classification accuracy of 96.33%, significantly outperforming traditional evaluation methods and offering a robust solution for pork freshness assessment.