Klasifikasi Mutu Telur Bebek Menggunakan Single Shot Multibox Detector Mobilenet

Abstract

Duck eggs are a source of easily digestible and highly nutritious animal protein. Duck eggs consist of 13% protein, 12% fat, and vitamins and minerals. Based on chemical properties, especially the nutritional value of eggs, it is highly recommended for consumption by children who are in their growth and development period, pregnant and lactating women, and the elderly. The advantages of duck eggs include being rich in minerals, vitamin B6, pantothenic acid, vitamin A, vitamin E, and vitamin B12. When selecting eggs, it's usually just to look at the outer shell of the egg and beat the egg. If the eggs don't sound when shaken, then the community considers the eggs are still good and fit to eat. But you can not only identify an egg from the outer shell, because the color of the outer shell doesn't necessarily look good, but you can ensure that the egg is of good quality and suitable for consumption. In this research, real-time detection and classification of duck egg quality was carried out using the Single Shot Multibox Detector Mobilenet algorithm. The data used in this study is input data which is divided into two types, namely test data and training data. The number of image data obtained is 120 images of duck eggs, of which 108 are training images and 12 are test images. From the experimental results it can be concluded that there are deviations when the learning rate values are varied in 11 trials. The average precision value of the confusion matrix is 71.32%, 73.79 precision when remembering is 60.95%, with the best precision value being 81.8% for learning rate values of 0.30.