Fruit Identification and Estimation of Calories Using YOLOv3 Algorithm

Calories are one of the most essential things health-conscious people consider when eating. Maintaining a good number of calories daily is vital for gaining energy and for the body to function correctly. This provides a chance to create a system that offers accurate calorie estimation depending on the system's detected image and the weight of the fruit. The system model, consisting of a custom dataset, was trained using the YOLOv3 algorithm. The system uses a Raspberry Pi camera module to take images and an HX711 load cell amplifier with a 5kg load cell to measure weight. Captured images and weights in grams undergo the identification and estimation of calories. As a result, the researchers can successfully detect fruit varieties of bananas and apples with an accuracy of 86%. The system’s statistical treatment between the calibrated weight sensor and a weighing scale shows no significant difference, showing accurate weight measurement. Thus, all objectives of the study were accomplished, allowing the researchers to determine the correct identification of the fruit and estimate its calories based on its weight.