생성형 AI 기반 이미지 변환 툴의 활용성과 한계성, 그리고 디자인 교육에의 함의 = The Possibilities and Limitations of Generative AI Image Conversion Tools and Their Implications for Design Education

국문 초록 (Abstract)

본 연구는 최근 등장한 AI 기반 이미지 생성 도구가 컴퓨터 디자인 과업수행에 미치게 될 영향과 디자인 교육에 대한 시사점을 논의하고자 한다. 이를 위해 본고는 첫째, AI 기반 이미지 생성 기술 네 가지인 생성적 적대 신경망(GANs), 디퓨전(Diffusion), 이미지 분할 기술(SAM), 그리고 대규모 언어모델(LLM)을 기반으로 구현된 여덟 가지 AI 디자인 툴(PhotoRestore, DragGAN, StyleDrop, Dreamfusion, Lasco.ai, Cesium JS, Pictory.ai, VisionPro)의 기능성을 분석하였다. 둘째, 이러한 AI 기반 디자인 툴이 디자인 유형별(2D 그래픽 디자인, 사용자 경험 디자인, 웹 디자인, 그리고 3D 모델링 및 애니메이션)로 어떤 활용성과 한계성을 가지는지 분석하였다. 분석결과 디자인 과업 수행 및 교육에 생성형 AI 기술이 도입되면 학습 자료의 시각적 표현을 통한 학습 효과 증진, 텍스트를 이미지로 쉽게 변환하여 복잡한 개념의 시각화, 3D 모델로의 이미지 변환을 통한 복잡한 개념의 깊은 이해, 그리고 동영상 이미지나 객체 생성 기술을 통한 가상현실의 활용 등에 매우 유용하게 활용될 수 있음을 알 수 있었다. 마지막으로, 이러한 생성형 AI 기술의 등장이 디자인 과업수행과 교육에 미치게 될 영향력과 유의사항에 대해 논의하였다.

다국어 초록 (Multilingual Abstract)

The primary objective of this study is to explore the potential applications of emerging AI-based image generation tools in various design domains and examine their implications for design education. To accomplish this, the paper begins by analyzing the functionalities of eight AI design tools (PhotoRestore, DragGAN, StyleDrop, Dreamfusion, Lasco.ai, Cesium JS, Pictory.ai, VisionPro) based on four types of AI-based image generation technologies: Generative Adversarial Networks (GANs), Diffusion, Segment Anything Model (SAM), and Large Language Models (LLM). Subsequently, the study investigates the practicality and limitations of these AI-based design tools in different design areas, such as 2D graphic design, user experience design, web design, and 3D modeling and animation. The analysis indicates that incorporating generative AI technologies into design tasks and education could significantly improve learning effectiveness through a visual representation of educational materials, facilitate the conversion of text into images for visualizing complex concepts, promote in-depth understanding of intricate ideas by transforming images into 3D models, and streamline the utilization of virtual reality through video image or object generation technology. Lastly, the paper discusses the potential impact and considerations that the advent of such generative AI technology may have on design tasks and education.

참고문헌 (Reference)

1 Ramesh, A, "Zero-shot text-to-image generation"

2 Brumberger, E., "Visual communication in the workplace : A survey of practice" 16 (16): 369-395, 2007

3 Wang, T. C., "Video-to-video synthesis" 1152-1164, 2018

4 Mercier, M., "Video Games Can Enhance Creativity: The Mediating Role of Psychological Capital"

5 Dove, G., "UX design innovation : Challenges for working with machine learning as a design material" 278-288, 2017

6 Cohen, N., "This is my unicorn, Fluffy: Personalizing frozen vision-language representations" Springer Nature Switzerland 558-577, 2022

7 Westerlund, M., "The emergence of deepfake technology: A review" 9 : 11-, 2019

8 Hoogendam, M., "The effect of digital tools on illustration" Delft University of Technology 2019

9 Edmonds, E., "The art of interaction: What HCI can learn from interactive art" Morgan & Claypool Publishers 2018

10 Kerlow, I. V., "The art of 3D computer animation and effects" Wiley 2004

11 Shadbolt, N., "The Semantic Web Revisited" 21 (21): 96-101, 2006

12 Floridi, L., "The 4th revolution: How the infosphere is reshaping human reality" Oxford University Press 2014

13 Ganin, Y, "Synthesizing programs for im- ages using reinforced adversarial learning"

14 Sohn, K., "StyleDrop:Text-to-Image Generation in Any Style"

15 Han Zhang, "StackGAN++: Realistic Image Synthesis with Stacked Generative Adversarial Networks" Institute of Electrical and Electronics Engineers (IEEE) 41 (41): 1947-1962, 2019

16 Pauly, M., "Shape modeling with point-sampled geometry" 22 (22): 641-650, 2003

17 Gadelha, M., "Shape Reconstruction from 3D and 2D Data Using P-NET" 13498-13507, 2020

18 Kirillov, A., "Segment anything"

19 Andrei, S. S., "SUPERVEGAN : Super resolution video en-hancement GAN for perceptually improving low bitrate streams" 9 : 91160-91174, 2021

20 Bowman, D. A., "Questioning naturalism in 3D user interfaces" 55 (55): 78-88, 2012

21 Chen, Z., "Pix2Vox: Context-aware 3D Reconstruction from Single and Multi-view Images" 2690-2698, 2019

22 Sabin, J., "Nonlinear systems–digital design and fabrication" 84 (84): 118-127, 2014

23 Marghitu, D., "Multimedia tools and applications" Springer 2004

24 Wu, J., "MarrNet : 3D Shape Reconstruction via 2. 5D Sketches" 34 (34): 13041-13048, 2020

25 McCloud, S, "Making comics: Storytelling secrets of comics, manga, and graphic novels" HarperCollins Publishers 2006

26 Mavridou, E., "Machine vision systems in precision agriculture for crop farming" 5 (5): 89-, 2019

27 Chen, W., "Learning to animate from real videos using recurrent mixture density networks"

28 Chen, Z., "Learning 3d shapes as multi-layered height maps with 2d convolutional networks" 33 : 4177-4184, 2019

29 Brock, A., "Large scale GAN training for high fidelity natural image synthesis"

30 Magnenat-Thalmann, N., "Joint-dependent local deformations for hand animation and object grasping" 26-33, 1988

31 Rogers, Y., "Interaction design: beyond human-computer interaction" John Wiley & Sons 2011

32 Dede, C., "Immersive interfaces for engagement and learning" 323 (323): 66-69, 2009

33 Gauthier, D., "Graphics for multi- media and the World Wide Web" Pearson Education 2017

34 Kingma, D. P, "Glow: Generative flow with invertible 1x1 convolutions" 10236-10245, 2018

35 Goodfellow, I., "Generative adversarial nets" 2672-2680, 2014

36 Vondrick, C., "Generating videos with scene dynamics" 613-621, 2016

37 Deng, S., "GAN-based automatic web design" 144-148, 2020

38 Sun, G., "Enhanced 3D Shape Reconstruction With Knowledge Graph of Category Concept" 18 (18): 1-20, 2022

39 Xu, B., "Empirical evaluation of rectified activations in convolutional network"

40 Poole, B., "Dreamfusion : Text-to-3D using 2D diffusion"

41 Pan, X., "Drag your GAN : Interactive Point-based manipulation on the generative image manifold"

42 Dhariwal, P., "Diffusion models beat GANs on image synthesis"

43 Gavin R. McCormack, "Differences in transportation and leisure physical activity by neighborhood design controlling for residential choice" Elsevier BV 8 (8): 532-539, 2019

44 Nielsen, J., "Designing web usability: The practice of simplicity" New Riders Publishing 2000

45 Huang, G., "Densely connected con- volutional networks" 4700-4708, 2017

46 Sohl-Dickstein, J, "Deep unsupervised learning using nonequilibrium thermodynamics" PMLR 2256-2265, 2015

47 Hsu, C. C., "Deep fake image detection based on pairwise learning" 10 (10): 370-, 2020

48 Chen, H., "Cross-modal image-text retrieval with se- mantic consistency" 1749-1757, 2019

49 Grossman, T., "Chronicle : Capture, Exploration, and Playback of Document Workflow Histories" 143-152, 2010

50 Elgammal, A., "CAN: Creative Adversarial Networks, Generating "art" by learning about styles and deviating from style norms"

51 Radu, I., "Augmented reality in education: A meta-review and cross-media analysis" 18 (18): 1533-1543, 2014

52 Xu, T., "AttnGAN: Fine-grained text to image generation with attentional generative adversarial networks" 1316-1324, 2018

53 Cowen, T., "Apple Vision Pro is receiving strong reviews"

54 Ethan Waisberg, "Apple Vision Pro and why extended reality will revolutionize the future of medicine" Springer Science and Business Media LLC 1-2, 2023

55 Garrett, N., "An e-portfolio design supporting ownership, social learning, and ease of use" 14 (14): 187-202, 2011