계산 이미지의 탄생 - 장치에서 알고리즘으로 -
The birth of computational image - from apparatus to algorithm -
0
조회
0
다운로드
2
인용
0
참고문헌
키워드
초록
This article deals with the changes triggered by artificial intelligence related to image production. The image originally started as a “manual image” drawn by human hands and then differentiated into a “technical image” produced by apparatus in the 19th century. Since then, with the development of image generation algorithms, the automaticity of image production has become more advanced. This generative model, represented by GAN, produces images only by 'calculation' based on probability and statistics. Currently, GAN is receiving the most attention in the field of computer vision, and it excels in various tasks such as creating high-resolution images, translating images, and compositing photos. GAN learns the original photos and creates an image similar to it, which goes beyond the limitations of apparatus such as human hands and cameras. This 'computational image' belongs to the category of 'technical image' in a broad sense, but differs from photography, the first technical image in that there is no indication object. In addition, the automaticity of the program, the nature of the apparatus, is accelerating the exclusion of humans from the image production process in an extreme form. As a result, humans cannot reflect their intentions in image production and are in a position to become a simple consumer of images.
Abstract
This article deals with the changes triggered by artificial intelligence related to image production. The image originally started as a “manual image” drawn by human hands and then differentiated into a “technical image” produced by apparatus in the 19th century. Since then, with the development of image generation algorithms, the automaticity of image production has become more advanced. This generative model, represented by GAN, produces images only by 'calculation' based on probability and statistics. Currently, GAN is receiving the most attention in the field of computer vision, and it excels in various tasks such as creating high-resolution images, translating images, and compositing photos. GAN learns the original photos and creates an image similar to it, which goes beyond the limitations of apparatus such as human hands and cameras. This 'computational image' belongs to the category of 'technical image' in a broad sense, but differs from photography, the first technical image in that there is no indication object. In addition, the automaticity of the program, the nature of the apparatus, is accelerating the exclusion of humans from the image production process in an extreme form. As a result, humans cannot reflect their intentions in image production and are in a position to become a simple consumer of images.
참고문헌 (23)
- Bergmann, Urs, "Learning Texture Manifolds with the Periodic Spatial GAN", ~ (2017)
- Besanon, "Alain. L’Image interdite: une histoire intellectuelle de l’iconoclasme" (1994), Fayard
- Brock, Andrew, "Large Scale GAN Training for Hige Fidelity Natural Image Synthesis", ~ (2019)
- Choi, Yunjey, "StarGAN:Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translation", ~ (2018)
- Flusser, Vilem, "Towards a philosophy of photography" (1983), Reaktion Books Ltd
- Flusser, Vilem, "Vilem(1985). Into the universe of technical image" (2011), University of Minnesota Press
- Goodfellow, Ian, "Generative Adversarial Nets", ~ (2014)
- Goodfellow, Ian, "NIPS 2016Tutorial: Generative Adversarial Networks", ~ (2017)
- Huang, Rui, "Beyond Face Rotation: Global and Local Perception GAN for Photorealistic and Identity Preserving Frontal View Synthesis", Proc. IEEE Int. Conf. Computer Vision, pp.2439~2448 (2017)
- Indolia, Sakshi, "Conceptual Understanding of Convolutional Neural Network – A Deep Learning Approach", Procedia Computer Science, 2018, pp.679~688 (2018)
- Isola, Phillip, "Image-to-Image Translation with Conditional Adversarial Networks", ~ (2018)
- Jetchev, Nikolay, "Texture Synthesis with Spatial Generative Adversarial Networks", ~ (2017)
- Karras, Tero, "Progressive Growing of GANs for improved quality, stability, and variation", ~ (2018)
- Kim, Taeksoo, "Learning to Discover Cross-Domain Relations with Generative Adversarial Networks", ~ (2017)
- Ledig, Christian, "Photo-Ralistic single image Super-Resolution using a Generative Adversarial Network", ~ (2017)
- Li, Chuan, "Precomputed Real-Time Texture Synthesis with Markovian Generative Adversarial Networks", ~ (2016)
- Mirza, Mehdi, "Conditional Generative Adversarial Nets", ~ (2014)
- O’Shea, Keiron, "An Introduction to Convolutional Neural Networks", ~ (2015)
- Pan, Zhaoqing, "Recent Progress on Generative Adversarial Networks (GANs): a survey", IEEE Access, 7, pp.36322~36333 (2019)
- Radford, Alec, "Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks", ~ (2016)
- Salimans, Tim, "Improved Techniques for Training GANs", ~ (2016)
- Zhang, Han, "Self-Attention Generative Adversarial Networks", Proceedings of the 36th International Conference on Machine Learning (2019)
- 박평종, "프로그램(Program)에서 포스트그램(Postgram)으로: 사진과 인공지능 예술의 근본문제", 현대미술학 논문집, 22(1), pp.107~130 (2018), 현대미술학회[DOI]