Automatic Measured Drawing Generation for Mokkan Using Deep Learning

Wataru Ohyama, Yoshinori Hatano, Hajime Baba

Abstract: This research proposes a method that automatically generates measured drawings for mokkan (wooden strips) using deep-learning technology. The proposed method inputs an image containing one strip of mokkan captured by smartphone or tablet cameras and outputs a measured drawing for the input image. This report details the proposed deep-learning-based method and the dataset collected for training and evaluation of the proposed method. The performance of the proposed method has been empirically confirmed using 337 images of real mokkan. The proposed method has several contributions that promote historical research of mokkan.