<![CDATA[Sasirangan is a traditional textile from South Kalimantan renowned for its unique motifs and deep cultural significance. However, the preservation of Sasirangan motifs is increasingly challenged by the declining number of skilled craftsmen and inadequate digital documentation. This study presents the development of an automated motif classification system to support the digital preservation of Sasirangan cultural heritage. The system was developed using the MobileNetV2 architecture with transfer learning from ImageNet pre-trained weights, implemented through the TensorFlow framework. A dataset comprising 70 images from 9 different Sasirangan motifs was utilized. To address the limited dataset size, various data augmentation techniques were applied. In the proof-of-concept phase, a binary classification task (Gigi Haruan vs. Unknown) was conducted using an 80:10:10 training-validation-test split. Experimental results demonstrated strong model performance, achieving 96.06% test accuracy for Gigi Haruan motif detection, 96.5% average F1-score, and 98.31% rejection accuracy for non-Sasirangan images. Additionally, a user-friendly web interface based on Gradio was developed, featuring real-time prediction through webcam integration. This study highlights the effectiveness of transfer learning in classifying traditional textile motifs and provides a solid foundation for future advancements, including multi-class classification and cloud-based database integration. The proposed system is expected to contribute significantly to the documentation, education, and preservation of Sasirangan cultural heritage in the digital era.]]>
