Identifikasi Kesegaran Ikan Bandeng Non-kontak menggunakan MobileNetV2

Milkfish is a superior commodity in several districts in Indonesia, namely Sidoarjo, Semarang, and Banten. This fish is also a favorite of Indonesians because it is nutritious and affordable. Therefore, for milkfish processed product business people, the freshness of milkfish is an important parameter because the freshness of the fish affects the quality of the processed products. Manual fish sorting is a problem when the number of fish is vast because it is prone to errors due to fatigue. In addition, manual fish sorting is also wasteful and time-consuming. Therefore, a non-contact automatic system is needed to identify fish freshness based on digital images. This study uses the Convolutional Neural Network (CNN) model to develop an application for milkfish freshness identification. We applied the MobileNetV2 model to identify the freshness of milkfish into three freshness classes, namely very fresh, fresh, and not fresh. The application uses the MobileNetV2 model on 312 milkfish images. The freshness classification performance reached 95%, 70%, and 80% in the high-fresh, fresh, and not-fresh classes, respectively. The global accuracy of the system reached 81.6%, indicating that the application can work well. From the experiments and analysis conducted, it can be concluded that the system has good capabilities in identifying fish freshness.