PENGELOMPOKAN KESEGARAN IKAN MELALUI CITRA MATA IKAN MENGGUNAKAN METODE CNN (Convolution Neural Network)

Abstract

The purpose of this study was to examine the Convolution Neural Network method for applying freshness levels of fish using fish eye images. Convolutional Neural Network (CNN) is a new model in the area of object recognition. Specifically for spatial data type input, CNN has a special layer, the convolution layer and the pooling layer that allows the hierarchical learning process of data features. The amount of data used for this study is 525 training data, 77 testing data, and 18 of test data with 9 categories, namely fish alive, dead on the first day, frozen 2 days, frozen 3 days, frozen 4 days, frozen 5 day, freeze 6 days, freeze 7 days, and rot. From 18 test data used there were 72,15% photos of fish eyes in accordance with the classification carried out and the highest epoch results displayed reached 98,44% accuracy.