Implementasi Metode Convolutional Neural Network Untuk Pengenalan Pola Motif Kain Tenun Rote Ndao Berbasis Android

Abstract

Rote ndao Ikat Weaving has interesting characteristics in each fabric motif because it has different motifs which indicate the ethnic differences contained in each of the resulting motifs. Rote Ndao weaving has a variety of motifs that are still unknown to many people, so in this study a classification of motifs of rote ndao woven fabrics was carried out using the Convolutional Neural Network method. Weaving motif classification uses 3 motifs with a total of 1050 data including 70 data for Ai Bunak, Dula Kakaik and Lafa Langgak motifs each. Data for 3 fabric motifs is divided into 80% training data and 20% testing data. The classification of the Rote Ndao Weaving motif is carried out by building an architectural model of MobileNetV2 plus Dropout and using a Learning Rate of 0.0003 which is trained and evaluated using K-Fold Cross Validation with a value of K=5, obtaining an accuracy of 93%; The precision is 94% for the Ai Bunak motif, 88% for the Dula Kakaik motif, and 100% for the Lafa Langgak motif. Recall of 87% for the Ai Bunak motif, 94% for the Dula Kakaik motif and 100% for the Lafa Langgak motif. Then the F-Score value obtained is 90% for the Ai Bunak motif, 91% for the Dula Kakaik motif and 100% for the Lafa Langgak motif.