Batik Banyumasan
Description
The 'Batik Banyumasan' dataset is a data set containing 5.148 images of typical batik motifs from Banyumas, a district in Central Java, Indonesia which is famous for its rich culture, including batik art. The data collected was in the form of digital images of Banyumasan batik with 11 different motifs, namely Angguran, Ayam Puger, Jahe Lumbon, Jahe Puger, Jahe Srimpang, Lumbon, Madu Bronto, Pring Sedapur, Puger Galar, Puger Telu Bal, and Wit Lumbon. This dataset was developed for research and development of pattern recognition technology, image analysis, image classification and preservation of Banyumasan batik culture.
Files
Steps to reproduce
The data collected were 572 images of Banyumasan batik motifs. The labeling of the image data was confirmed directly by Mr Heru Santoso the owner of Rumah Batik R so that it can be accounted for. Image data was collected using a camera with specifications of 64MP Wide Angle Camera, sensor size 1/1.97", pixel size 0.7 μm, and Aperture f/1.79. Camera settings use features such as F1/8, shutter speed 1/33, ISO 250, focal length 5 mm, white balance auto, and no flash. Batik motifs were captured one by one to capture the details of their texture. Photos were taken by spreading the batik cloth on the floor with a camera distance of 50 cm from the batik cloth with a top-angle camera placement. In addition, photos were taken during the day in open room conditions, not exposed to direct sunlight, and no lamp light, so the light intensity was around 10,000 to 20,000 lux. After the image data was collected, an augmentation process was carried out on each image data. This aims to increase the quantity and diversity of data used in model training. The augmentation process uses one of the Python libraries, namely Albumentations. There are eight image data augmentation parameters used, namely channel shuffle, elastic transform, flip, fourier domain adaptation, random grid shuffle, random rotate 90, sharpen, and transpose. Each image that has been augmented produces eight new augmented images. The number of datasets that previously amounted to 572 images, after the augmentation process became 5,148 images.