Implementasi Sistem Deteksi dan Klasifikasi Makanan untuk Estimasi Kandungan Kalori menggunakan YOLOv11 dan EfficientNetB3

Abstract

As public awareness of healthy eating habits increases, the demand for nutritional information becomes increasingly essential. However, access to food calorie content remains limited, especially for homemade or unlabeled meals. This study proposes the development of an Android-based application capable of automatically detecting and classifying food from images, as well as estimating its calorie content. The system integrates the YOLOv11 algorithm for object detection and EfficientNetB3 for food classification, with calorie data sourced from the trusted FatSecret Indonesia. The Dataset includes 10 local food categories and was processed through annotation, augmentation, and model training using TensorFlow. The evaluation results show that the YOLOv11 detection model achieved a Precision of 87,7%, Recall of 84.5%, and mAP@50 of 89,6%. Meanwhile, the EfficientNetB3 classification model achieved an accuracy of 92%, Precision of 92%, Recall of 92%, and an F1-Score of 91%. The system was implemented into an Android application by converting the model into TensorFlow Lite format. This system allows users to obtain calorie estimates conveniently through food images, thus assisting in daily calorie intake monitoring.