딥러닝에 기반한 전통성의 비율별 한옥카페 디자인 가이드라인

Abstract

This paper presents a direction that can be used as a reference for design by classifying images according to the ratio of traditionality of spaces that appear in hanok cafes. With the development of the times, the traditional hanok, which felt uncomfortable, was recklessly repaired and remodeled without standards to create a space suitable for the times. Therefore, in this paper, for the universalization of hanok in the modern era, the purpose of this paper is to examine the architectural elements by dividing the traditionality ratio of the hanok cafe by focusing on the hanok cafe, which is being used because of the transition of space rather than the hanok space, which is the meaning of residence. As a procedure for such research, learning data images of traditional hanok and modern interiors were collected and preprocessed. In addition, the test image was selected as the integrated result of the image data classified through prior research and the image data classified through deep learning, and Grad-CAM was used to check the distribution of traditionality. In deep learning, most of the ceiling wood is recognized as traditional, and the column is recognized in the structural part. And most of the furniture was perceived as a modern interior, and when low furniture made of wood was used, it appeared in the proportion of a traditional hanok. On the walls, traditional hanok was finished with plaster, clay, or Hanji(Korean paper), but in modern interiors, white walls were finished, so it was not recognized as a part that showed a lot of tradition. If the traditional ratio is low, even if the overall atmosphere leads to a modern atmosphere if a traditional ratio on the ceiling, create the atmosphere of a traditional hanok cafe.