RESTORE THE PURITY IN ART THROUGH ARTIFICIAL INTELLIGENCE
Dansaekhwa refers to a group of monochrome paintings that represented Korean abstract painting of the 1970s and have emerged again after 40 years. Credited as one of the proper nouns in the global art scene, dansaekhwa has successfully strengthened its position. Drawing worldwide attention in a short period of time, dansaekhwa is laden with Korea’s intrinsic emotion and history remarkably distinguishable from those aroused by Western minimalism as well as considered to be contemporary Korean art’s freestanding brand. Therefore, we have to first of all understand the social phenomena of the 1970s when dansaekhwa first emerged in order to grasp this painting genre properly.
Any cardinal art tendency tends to come into being inextricably bound up not only with art history and aesthetics but also political, economic, and social aspects. The monochrome of the 1970s often referred to as “the white monochrome” and white was a color suggestive of Korean people’s emotion and racial characteristics. Yusuke Nakahara denoted the unique hallmarks of Korean dansaekhwa in the catalog essay for Five Korean Artists, Five Kinds of White, a group show featuring Korean dansaekhwa held at Tokyo Gallery in 1975. Art critic Lee Yil discovered the possibility that Korean dansaekhwa could be defined collectively with the single term “Korean monochrome,” thereby striving to spread it to the entire art scene. In the 1970s, however, artists called monochrome painters did not use this term actively and only a few of artists in the Korean art community gave prominence to this painting genre.
In the 2010s, dansaekhwa was revived amid a globalized economic recession and a depressed art market. This was a sort of the movement to revive Korean art in a continuing economic recession. Unlike monochrome paintings of the 1970s, dansaekhwa in this period discarded the symbolism of white but works by painters commonly called “dansaekhwa artists” had something in common: paintings done in this period were made up of colors that looked visually similar and in sync with one another. Each artist has their own distinctive methodological traits and idioms. As its appellation implies, however, there are elemental limits. As mentioned above, dansaekhwa cannot depart from the conceptual limits of monochrome paintings or those with very limited colors and tones.
Dansaekhwa has been marketed and promoted through an emphasis on “Koreanness” or “Korean identity.” Most works of dansaekhwa are obviously marked by their collective trait, considering only this type of painting itself without involving any other factors. However, no academic issues and questions, such as how this mode of painting represents Korean identity and what invariable Korean identity is despite a time difference of 40 years, have not been fully discussed. Some fears that dansaekhwa has been established based on a feeble theoretical foundation and may have encouraged lopsided preferences in that abstract paintings by senior artists have monopolized the market. Although dansaekhwa became a sensation in the global art scene of the 2010s, its boom was enjoyed at only a few galleries and by very few artists as its artists and works were extremely limited. The majority of artists and the public were not involved in this fever. Therefore, there is a critical evaluation that dansaekhwa was exploited as the means to bring about an abnormal concentration of wealth to a small minority of artists, excluding the majority of people who represent Korean identity.
The project is intended to explore the relationship between artwork, creator, and observer as well as the true essence of art. Dansaekhwa is selected as a subject of the project particularly for its relevancy to the current Korean art scene, and global acclamation. The major aims for the project are:
1. to devise a method to generate the images that are similar to the original artworks with minimal human intervention.
2. to evaluate the meaning of the synthesized images, generated with the absence of the traditional artists.
3. to re-examine the role/significance of human creators in the art-making process
The code used in the project are heavily borrowed from ArtGAN (https://github.com/cs-chan/ArtGAN/tree/master/ArtGAN). ArtGAN proposes a novel framework to synthetically generate more challenging and complex images, which is in contrast to most of the current solutions that focused on generating natural images such as room interiors, birds, flowers and faces. The model suggests innovative approaches for conditional image synthesis, allowing backpropagation of the loss function with respect to the labels to the generator from the discriminator. As a result, ArtGAN exhibits the phase of training that is dramatically less susceptible to mode collapse compared to other traditional models.
The four series of Dansaekhwa are chosen for their formal appropriacy and the amount of data available. The selected four are "From Line선으로부터" by Ufan Lee, "Ecriture묘법" by Seobo Park, "Correspondence조응" By Ufan Lee, and "Untitled무제" by Yun Hyong-keun. Based on over 12,000 image files achieved through web crawling, and data augmentation, the four separate datasets were created respectively.
Examples of the selected images. From top to bottom: From Line, Ecriture, Correspondence, and Untitled
The major changes to the vanilla ArtGAN are the use of customized data loader function to facilitate a more general use, and the enlarged output size. After applying parameter tuning and network reconfiguration, the revised model stably generate the images with a size of 256 X 256 pixels, compared to the existing model, which generates 128 X 128-pixel images.
Examples of the synthesized images. From top to bottom: From Line, Ecriture, Correspondence, and Untitled
After applying parameter tuning and network reconfiguration, the revised model stably generate the images with a size of 256 X 256 pixels, compared to the existing model, which generates 128 X 128-pixel images. As the above samples indicate, the synthesized images exhibit formal features (e.g. shapes, colors, textures, brushstrokes, and etc.) that are similar to the original images.
The last one-third of training. The model exhibits an unconventional way of learning
This project started with the premise that any art or trends in art cannot be separated from the current social conditions. As reviewed above, however, dansaekhwa fever was developed divorced from the political and social realities the general public faced.
Dansaekhwa has not only promoted prejudiced taste but also used as a means to increase the wealth of the few prosperous investors. Under the influence of capitalism, it is not unreasonable to regard a work of art as goods, real properties, or investments. Unfortunately, as we have been observing with this very case of dansaekhwa, there are many problems arise when artworks are thought merely as commercial goods.
Art forgeries, ghostwriter artists, aggressive marketing, and lack of supporting academic, and scholastic research are all issues surrounding dansaekhwa and most of which are due to its abnormally high market demand. With this current tendency in mind, by eliminating the presence of human artists, which have fallen as a mere brand name, in the creation of dansaekhwa paintings, the project aims to contemplate on the true essence of art, which should not be discolored even in capitalistic society. The project acknowledges the involvement of unconventional artist in the artmaking process. By minimizing human intervention, the project suggests that the future potentials of artificial intelligence as a tool to restore purity in art and to challenge the traditional definition of creative practice.
Tan, W. R., Chan, C. S., Aguirre, H. E., & Tanaka, K. (2017). ArtGAN: Artwork synthesis with conditional categorical GANs. 2017 IEEE International Conference on Image Processing (ICIP). doi:10.1109/icip.2017.8296985