Enhancement of traditional clothes pattern recognation using Convolutional Neural Network

In general, Traditional clothes in Indonesia has its own characteristics to recognize the origin of the clothes. The formation of traditional clothes patterns in Indonesia from the crafts of remote communities in the regions of Indonesia. The article is that Indonesian citizens still find it difficult to distinguish traditional clothes that are scattered throughout Indonesia. To support this research, we use the Convolutional neural network method for types VGG16, VGG19 and MobileNetV2. To support the recognition process for a variety of traditional clothes patterns, the pattern processing process has started from preprocessing and image segmentation. This research was conducted on 5 types of experiments with supporting parameters, namely the number of training images and test images, rotation and scale of each image, image class and CNN parameters. After we conducted experiments on a collection of images that could be found in various scales and degrees, the average classification accuracy using VGG 16 was 79,23% in condition 2, but the highest classification accuracy was between 84,2%-92,5%. Meanwhile, if using VGG 19 the average classification accuracy is 79,95% with condition 1 for the highest accuracy between 92,6%-96,3%. If using mobilenetV2 the average classification accuracy is 83,44% for the highest accuracy between 87,3%-96,4%. Basically Research for traditional clothes pattern recognition using Deep CNN has succeeded in recognizing some traditional clothes patterns up to more than 92%.