Food Image Classification for Maternal Nutritional Fulfillment Using MobileNet

Proper nutrition during pregnancy is crucial for the health of both the mother and the fetus. However, many pregnant women struggle to maintain an adequate diet due to a lack of accessible dietary monitoring tools. This study aims to enhance dietary monitoring for pregnant women by comparing the performance of traditional Convolutional Neural Networks (CNN) and the pre-trained MobileNet V2 model in food image classification. Given the increasing reliance on mobile health applications, especially in regions like Indonesia, it is essential to develop efficient and accurate food recognition systems that can operate on mobile devices. Our research involves training and evaluating both CNN and MobileNet V2 models on a dataset of food images commonly consumed by pregnant women. The models' performance is assessed based on accuracy, precision, recall, and inference time. The results show that MobileNet V2 significantly outperforms the traditional CNN model, achieving accuracy, recall, precision, and F1-score of $0.83,0.82,0.85$, and 0.85 respectively, compared to $0.61,0.60,0.62$, and 0.64 for the CNN model. These findings highlight MobileNet V2’s superior efficiency and accuracy, making it ideal for real-time dietary monitoring on mobile devices. Future research should focus on integrating additional features into the dataset and exploring advanced optimization techniques to further enhance model performance and generalizability, ultimately contributing to better maternal and fetal health outcomes.