PENERAPAN ALGORITMA CONVOLUTIONAL NEURAL NETWORK (CNN) RESNET-50 PADA SISTEM KLASIFIKASI MOTIF KAIN TENUN TEMBE NGGOLI BIMA DONGGO
Kata Kunci:
Convolutional Neural Network, Resnet-50, Transfer Learning, Klasifikasi Citra, Tembe Nggoli Bima Donggo, CRISP-DMAbstrak
Motif kain tenun Tembe Nggoli Bima Donggo merupakan salah satu warisan budaya yang memiliki beragam pola dan karakteristik visual. Proses klasifikasi motif masih dilakukan secara manual sehingga membutuhkan ketelitian, waktu, dan pengetahuan khusus. Penelitian ini bertujuan untuk membangun model klasifikasi motif kain tenun Tembe Nggoli Bima Donggo menggunakan metode Convolutional Neural Network (CNN) dengan arsitektur ResNet-50 sehingga proses klasifikasi motif dapat dilakukan secara otomatis. Metode penelitian yang digunakan adalah Cross Industry Standard Process for Data Mining (CRISP-DM), yang meliputi tahapan business understanding, data understanding, data preparation, modeling, evaluation, dan deployment. Dataset yang digunakan terdiri dari 1.785 citra yang terbagi ke dalam 22 kelas motif. Seluruh citra melalui proses preprocessing berupa penyeragaman ukuran menjadi 224 × 224 piksel, kemudian diterapkan data augmentation menggunakan teknik rotasi, horizontal flip, zoom, width shift, height shift, brightness, dan shear. Model dibangun menggunakan pendekatan transfer learning dengan ResNet-50 sebagai base model, kemudian dilakukan proses fine-tuning pada 30 lapisan terakhir untuk meningkatkan kemampuan model dalam mengenali karakteristik motif. Hasil pengujian menunjukkan bahwa model berhasil mencapai nilai accuracy sebesar 97,00%, precision sebesar 98,05%, recall sebesar 97,00%, dan F1-score sebesar 97,08%. Hasil tersebut menunjukkan bahwa model ResNet-50 mampu mengklasifikasikan motif kain tenun Tembe Nggoli Bima Donggo dengan tingkat ketepatan yang sangat baik. Penelitian ini diharapkan dapat menjadi alternatif dalam membantu proses klasifikasi motif kain tenun secara lebih cepat, konsisten, dan objektif serta mendukung upaya pelestarian budaya melalui pemanfaatan teknologi deep learning.
Tembe Nggoli Bima Donggo woven fabric motifs represent one of Indonesia's cultural heritages, characterized by diverse patterns and distinctive visual features. The classification of these motifs is still performed manually, requiring considerable time, accuracy, and specialized knowledge. This study aims to develop an automatic classification model for Tembe Nggoli Bima Donggo woven fabric motifs using the Convolutional Neural Network (CNN) method with the ResNet-50 architecture. The research employed the Cross Industry Standard Process for Data Mining (CRISP-DM), consisting of business understanding, data understanding, data preparation, modeling, evaluation, and deployment. The dataset consisted of 1,785 images representing 22 motif classes. All images were preprocessed by resizing them to 224 × 224 pixels, followed by data augmentation techniques including rotation, horizontal flip, zoom, width shift, height shift, brightness adjustment, and shear transformation. The classification model was developed using a transfer learning approach with ResNet-50 as the base model, followed by fine-tuning of the last 30 layers to improve its ability to recognize motif characteristics. The experimental results demonstrated that the proposed model achieved an accuracy of 97.00%, precision of 98.05%, recall of 97.00%, and F1-score of 97.08%. These results indicate that the ResNet-50 model is highly effective in classifying Tembe Nggoli Bima Donggo woven fabric motifs. This research is expected to provide an alternative approach for faster, more consistent, and objective motif classification while supporting the preservation of traditional woven fabric through the application of deep learning technology


