Klasifikasi Kesegaran Ikan Lemuru Melalui Citra Mata Menggunakan Metode CNN Dengan Arsitektur MobileNetV2

Fish freshness is a major factor in determining the quality and marketability of fish. One of the main indicators used to assess fish freshness is eye condition. In several previous studies, the Convolutional Neural Network (CNN) method has been used to classify fish freshness based on eye images, with various architectures showing different performance. This study aims to develop a freshness classification model of lemuru fish through eye image using CNN with MobileNetV2 architecture. MobileNetV2 was chosen because it has a lightweight and efficient structure, enabling fast classification without sacrificing accuracy. The main differences applied include the addition of the lemuru eye image dataset and the augmentation technique applied. The dataset of this study consists of primary and secondary datasets. The secondary dataset consists of 480 images (240 fresh, 240 non-fresh) obtained from previous research. Meanwhile, the primary dataset consists of 72 images (36 fresh, 36 not fresh) collected directly from the Fish Auction Place (TPI) in Jembrana, Bali. The test results showed that MobileNetV2 with the addition of the primary dataset and the application of augmentation techniques achieved 100% accuracy using the test data. This test result shows that the MobileNetV2 model performs better than VGG16 in the research of lemuru freshness classification through eye image.