<![CDATA[Weaving is a cultural product that reflects the identity of the people who make it, with each region having its patterns, beauty, and distinctive features of its weaving motifs. However, identifying the origin of the region based on woven fabric motifs is often difficult to do due to the unique and diverse characteristics of the motifs. This paper aims to evaluate the performance of the MobileNetV2 architectural model in classifying the motif image of Sumbawa woven fabrics. This model was tested using a dataset of woven fabric images that included various motifs from Sumbawa. The results showed that the model managed to achieve the highest accuracy of 98.14% in the 20th and 25th epochs, with a training time of less than 1 hour. In the training data, the model obtained an accuracy of 99.71% with a loss of 12.99%, which indicates that the model can recognize images with a very high level of accuracy. However, in the validation data, the accuracy of the model was recorded at 92.71% with a loss of 41.98%, which shows that despite the decrease in accuracy, the model is still able to generalize well on data that has never been encountered before. In addition, the model showed excellent results in terms of precision (98.14%), recall (100%), and f1-score (99%). These findings confirm the effectiveness of the MobileNetV2 model in recognizing Sumbawa woven fabric motifs and provide a solid basis for the development of an automated system in supporting the preservation and promotion of regional weaving culture. This paper also shows the importance of model optimization to improve accuracy on validation data and reduce the gap between training data and unseen data. As a next step, the research can be directed to expand the dataset with more variations of motifs and regions to improve the model's ability to generalize to different types of woven fabric motifs.]]>
