Apry Aditya Saputra, Khakam Ma'ruf, Rizal Justian Setiawan, Nur Evirda Khosyiati, Nur Azizah
In today's era, public awareness of the importance of maintaining health and a balanced diet has increased significantly. Therefore, technological solutions are needed to support dietary monitoring through tools that are already part of everyday life. In accordance with this objective, this study developed a deep learning-based food detection and calorie estimation system integrated into a mobile application using MobileNetV3 transfer learning and OpenCV. The system was designed to enable real-time and accurate dietary monitoring as an alternative to traditional manual methods. The Food-101 dataset, containing 101,000 images across 101 food classes, was used to train two configurations of the MobileNetV3-Large model. Model 1 employed a dense layer architecture of 512 → 256 neurons with the RMSprop optimizer, while Model 2 adopted a 256 → 256 neuron configuration with the Adam optimizer. Experimental results showed that Model 1 achieved a testing accuracy of 61.79%, outperforming Model 2, which reached 55.72%. Although both models exhibited overfitting, Model 1 demonstrated stronger generalization ability. Integration with OpenCV enabled real-time image preprocessing within the mobile application, while the Calories Labels dataset provided calorie information for each recognized food item. The system was successfully deployed on the Android platform with a userfriendly interface, supporting digital health monitoring and the tracking of users' daily calorie intake. © 2025 IEEE.
Universitas Gadjah Mada, Faculty of Engineering, Dept of Electrical Engineering and Information Technology, Yogyakarta, Indonesia; Universitas Gadjah Mada, Faculty of Engineering, Dept of Mechanical and Industrial Engineering, Yogyakarta, Indonesia; National Chung Hsing University, College of Law and Politics, Dept of Asia and China Studies, Taichung, Taiwan; Universitas Negeri Yogyakarta, Faculty of Engineering, Dept of Culinary Engineering Education, Yogyakarta, Indonesia; National Cheng Kung University, College of Medicine, Dept of Public Health, Tainan, Taiwan
