With the advent of modernism, trends in art have shifted toward a greater autonomy in the definition, and practice of art. Art is no longer bounded by the constraints of traditional conceptions of beauty, and more importantly, of its primary function of representation. The result of such trends is, unfortunately, the widening distance between the culture of the art connoisseurs, and that of the larger public. Especially in today's highly modernized, and capitalist society, where fast gratification is widely available with the advanced technologies, art is often perceived as something esoteric, or something irrelevant to one's daily life. At this confusing time in the history of art, there seems to be no right answer to how art or the experience of art should be defined.
One may suggest that aesthetic experience in response to works of art should be defined as an experience qualitatively different from everyday experience. It is apparent that works of art can allow people the feelings of admiration, fascination, and awe, which are certainly distinguishable from everyday emotions; however, the value of art would truly manifest only when it is enjoyed as a part of our daily lives. Among its countless functions, one of the irremovable functions of art is to be consumed, and to be enjoyed, as if it is nearly impossible to ponder the meaning of art without considering the presence of its audience or appreciator.
But still there remain some big questions, such as about what to appreciate, or what to qualify as a work of art. Despite the social notion on the judgment of taste, which has been reinforced by the gap between the two cultures, of the art experts and of the public at large, there is no innate hierarchical measure in judgments of taste, or of art itself. The efforts, however, have been made in determining the value of works of art, either for matters of convenience or sometimes for one's benefit. With the rise of the art market, the works of art are typically being evaluated according to the two separate terms: an aesthetic, and a commercial value. Although setting the price for artwork, the product of human gift, might be considered as an attempt to degrade human creativity, the commercial value is now an inseparable part of the artwork as a whole, indirectly representing people's demand in art.
Nowadays, the works of art that are preferred by the larger public appear to be the works that can offer unquestionable beauty without invoking any complex philosophical questions. Such artworks are produced with clear objectives to fulfill human desire for beauty, functioning as an instrument to offer visual satisfaction. Among those works, some observable similarities can be found as listed below.
- There are (preferably) no discernible figures or shapes;
- The mood or atmosphere of an artwork provides positive (or at least neural) connotations;
- Any negative reactions (e.g. fear, contempt, disgust, and etc.) should not be evoked to the viewers
If what has been presented exceeds the underlying aesthetic or philosophical intention in terms of significance, the mere appearance of an artwork then becomes the primary reason for its existence. In the production of such kind of works, the presence of the creators or artists loses its importance. And the potential of artificial intelligence as an art-maker begins to reveal at this very point.
THE PROJECT
The project had been proceeded with the aims to shed light on the true value of art, and to challenge the traditional conceptions of art. 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 major change to ArtGAN is the use of customized data loader function to facilitate a more of general use. The new label Abstract Art has been added to the original Wikiart dataset. Abstract Art contains 4,000 images (1,000 collected from web sources, 3,000 data augmented) that qualify the above-listed characteristics.
Sample of the real artwork
Sample of the generated artwork
Although it should be acknowledged that there still are numerous limitations of an intelligent agent in the creative field of art, the results have shown that the images generated by artificial agents can elegantly fulfill one of the pivotal functions of art, offering an unconditional visual satisfaction to the viewers. The generative model is also capable of generating a massive number of images. Thus, despite the absence of the traditional artists, the images generated by ArtGAN, and by other generative models, have revealed their future potentials to satisfy the larger demographics of people in terms of appearance and a method of production.
With the advent of modernism, trends in art have shifted toward a greater autonomy in the definition, and practice of art. Art is no longer bounded by the constraints of traditional conceptions of beauty, and more importantly, of its primary function of representation. The result of such trends is, unfortunately, the widening distance between the culture of the art connoisseurs, and that of the larger public. Especially in today's highly modernized, and capitalist society, where fast gratification is widely available with the advanced technologies, art is often perceived as something esoteric, or something irrelevant to one's daily life. At this confusing time in the history of art, there seems to be no right answer to how art or the experience of art should be defined.
One may suggest that aesthetic experience in response to works of art should be defined as an experience qualitatively different from everyday experience. It is apparent that works of art can allow people the feelings of admiration, fascination, and awe, which are certainly distinguishable from everyday emotions; however, the value of art would truly manifest only when it is enjoyed as a part of our daily lives. Among its countless functions, one of the irremovable functions of art is to be consumed, and to be enjoyed, as if it is nearly impossible to ponder the meaning of art without considering the presence of its audience or appreciator.
But still there remain some big questions, such as about what to appreciate, or what to qualify as a work of art. Despite the social notion on the judgment of taste, which has been reinforced by the gap between the two cultures, of the art experts and of the public at large, there is no innate hierarchical measure in judgments of taste, or of art itself. The efforts, however, have been made in determining the value of works of art, either for matters of convenience or sometimes for one's benefit. With the rise of the art market, the works of art are typically being evaluated according to the two separate terms: an aesthetic, and a commercial value. Although setting the price for artwork, the product of human gift, might be considered as an attempt to degrade human creativity, the commercial value is now an inseparable part of the artwork as a whole, indirectly representing people's demand in art.
Nowadays, the works of art that are preferred by the larger public appear to be the works that can offer unquestionable beauty without invoking any complex philosophical questions. Such artworks are produced with clear objectives to fulfill human desire for beauty, functioning as an instrument to offer visual satisfaction. Among those works, some observable similarities can be found as listed below.
- There are (preferably) no discernible figures or shapes;
- The mood or atmosphere of an artwork provides positive (or at least neural) connotations;
- Any negative reactions (e.g. fear, contempt, disgust, and etc.) should not be evoked to the viewers
If what has been presented exceeds the underlying aesthetic or philosophical intention in terms of significance, the mere appearance of an artwork then becomes the primary reason for its existence. In the production of such kind of works, the presence of the creators or artists loses its importance. And the potential of artificial intelligence as an art-maker begins to reveal at this very point.
THE PROJECT
The project had been proceeded with the aims to shed light on the true value of art, and to challenge the traditional conceptions of art. 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 major change to ArtGAN is the use of customized data loader function to facilitate a more of general use. The new label Abstract Art has been added to the original Wikiart dataset. Abstract Art contains 4,000 images (1,000 collected from web sources, 3,000 data augmented) that qualify the above-listed characteristics.
Sample of the real artwork
Sample of the generated artwork
Although it should be acknowledged that there still are numerous limitations of an intelligent agent in the creative field of art, the results have shown that the images generated by artificial agents can elegantly fulfill one of the pivotal functions of art, offering an unconditional visual satisfaction to the viewers. The generative model is also capable of generating a massive number of images. Thus, despite the absence of the traditional artists, the images generated by ArtGAN, and by other generative models, have revealed their future potentials to satisfy the larger demographics of people in terms of appearance and a method of production.
DANSAEKHWA GENERATOR
IMAGE TRANSFORMATION TASK ON THE DANSAEKHWA DATASETS
Technological advances have brought tremendous changes not only to our general lives but also to the domain of art. As technology continues to progress, it has been functioning as a provider of new tools and applications for the arts and design disciplines. Among myriad applications, one of the trending topics is image transformation tasks, such as image-to-image translation, and style transfer.
Style transfer conventionally refers to the technique of reconstructing images in the style of other images. With the introduction and development of neural stylization methods, it is now possible to effectively, and artistically transfer input images into the images of any desired styles. The technique has allowed people who do not have any prior knowledge in art or coding to confidently produce artistic images that they desire to create. Various applications of style transfer are now available, ranging from the CycleGAN, a variant of GANs (Generative Adversarial Networks), to Prisma, a popular photo-editing app.
Closely related to neural style transfer, pix2pix is a classic example of image-to-image translation networks. These networks use a conditional generative adversarial network (cGAN) to learn a mapping from an input image to output image. The networks also learn a loss function to train the mapping, thus demonstrating a more effective approach to solving image-to-image translation tasks compared to other traditional models.
The images below are the examples of the image-to-image translation tasks available with pix2pix:
Retrieved from the paper Image-to-Image Translation with Conditional Adversarial Nets by Isola et al.)
THE PROJECT
The project was conducted as a part of the Dansaekhwa Project with the particular objectives of pondering the meaning of a creator in the artistic practice. The code used in the project were heavily borrowed from pix2pix-tensorflow (https://github.com/affinelayer/pix2pix-tensorflow). Among several options provided by the Dansaekhwa members, "Correspondence조응", the painting series by Lee Ufan, and "Untitled무제", a unique Umber Blue painting series, by Yun Hyong-keun are selected for their formal appropriacy, a perceivable similarity between each painting in the collections.
Examples of the target (original) images from the Untitled series by Yun Hyong-keun.
Examples of the target (original) images from the Correspondence series by Lee Ufan.
Each dataset consists of 500 pairs of the images, of 400 pairs used for training and the rest used for the test, which have a dimension of 512 × 512 pixels. 1,000 handmade equivalent input images for target images were created for these particular translation tasks. The images below are the examples of the pairs in each dataset.
The major issue when implementing the vanilla pix2pix was the pronounced checkerboard artifacts, one of the hallmarks of the artificially synthesized images. In order to alleviate the artifacts, the code has been added to the original model. The inserted code was referenced from Deconvolution and Checkerboard Artifacts, the article by Odena et al.
The pronounced checkerboard artifacts in the images generated by the original model. Images partially enlarged.
After inserting the code to resize convolution layers using bicubic resampling, the model stably generates more naturalistic images without exhibiting the pronounced checkerboard pattern of artifacts. The two collections of images below are the selection of output images comparing the results generated by the vanilla pix2pix and by the revised pix2pix networks.
Comparison of generated images with (bottom) and without (top) applying bicubic interpolation.
Comparison of generated images with (bottom) and without (top) applying bicubic interpolation.
CONCLUSION
As the above results of image-to-image translation tasks with pix2pix have revealed, the development of the image transformation technology has made it possible for people to create artistic images with minimal guidelines. Thus, the potential of artificial intelligence lies in its role as a collaborator, not there to replace human presence but to assist them to create something new, and something beautiful in a way that is not comparable with that of the past. The engagement of artificial intelligence in the domain of art has brought huge differences in our way of thinking, and in the perception of art and such change leads to the development of the new aesthetics in the digital world. Along with the emergence of altered perception, and of new methods of making art, human creativity will bloom elegantly with the help of artificial intelligence.
References
Zhu, J., Park, T., Isola, P., & Efros, A. A. (2017). Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks. 2017 IEEE International Conference on Computer Vision (ICCV). doi:10.1109/iccv.2017.244
Isola, P., Zhu, J., Zhou, T., & Efros, A. A. (2017). Image-to-Image Translation with Conditional Adversarial Networks. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). doi:10.1109/cvpr.2017.632