<feed xmlns="http://www.w3.org/2005/Atom"> <id>https://donghweeyoon.github.io/</id><title>Jeremy's Blog</title><subtitle>Jekyll + Chirpy theme powered by GitHub Pages</subtitle> <updated>2025-10-13T10:58:18+09:00</updated> <author> <name>Donghwee Yoon</name> <uri>https://donghweeyoon.github.io/</uri> </author><link rel="self" type="application/atom+xml" href="https://donghweeyoon.github.io/feed.xml"/><link rel="alternate" type="text/html" hreflang="ko-KR" href="https://donghweeyoon.github.io/"/> <generator uri="https://jekyllrb.com/" version="4.4.1">Jekyll</generator> <rights> © 2025 Donghwee Yoon </rights> <icon>/assets/img/favicons/favicon.ico</icon> <logo>/assets/img/favicons/favicon-96x96.png</logo> <entry><title>[Review] InfinityGAN: Towards Infinite-pixel Image Synthesis</title><link href="https://donghweeyoon.github.io/posts/infinitygan/" rel="alternate" type="text/html" title="[Review] InfinityGAN: Towards Infinite-pixel Image Synthesis" /><published>2022-08-21T18:02:19+09:00</published> <updated>2022-08-21T18:02:19+09:00</updated> <id>https://donghweeyoon.github.io/posts/infinitygan/</id> <content type="text/html" src="https://donghweeyoon.github.io/posts/infinitygan/" /> <author> <name>Donghwee Yoon</name> </author> <category term="AI" /> <category term="Paper Review" /> <summary>Lin, C. H., Lee, H. Y., Cheng, Y. C., Tulyakov, S., &amp;amp; Yang, M. H. (2021, September). InfinityGAN: Towards Infinite-Pixel Image Synthesis. In International Conference on Learning Representations. 서론 무한대 해상도의 영상을 합성할 수 있는 모델을 설계할 수 있을까? 저자는 InfinityGAN을 통해 위 물음에 답하고자 했다. 사실 InfinityGAN은 무한대 해상도의 영상을 합성할 수 있는 최초의 모델은 아니다. 무한대 해상도의 영상을 합성할 수 있는 모델들은 이미 제안된 적이 있다[1, 2]. 하지만 모델들의 기능은 텍스쳐와 같이 반...</summary> </entry> <entry><title>[Review] Towards faster and stabilized GAN training for high-fidelity few-shot image synthesis</title><link href="https://donghweeyoon.github.io/posts/fastgan/" rel="alternate" type="text/html" title="[Review] Towards faster and stabilized GAN training for high-fidelity few-shot image synthesis" /><published>2022-07-18T16:17:40+09:00</published> <updated>2022-07-18T16:17:40+09:00</updated> <id>https://donghweeyoon.github.io/posts/fastgan/</id> <content type="text/html" src="https://donghweeyoon.github.io/posts/fastgan/" /> <author> <name>Donghwee Yoon</name> </author> <category term="AI" /> <category term="Paper Review" /> <summary>Bingchen Liu, et al., “Towards faster and stabilized GAN training for high-fidelity few-shot image synthesis”, ICLR, 2021 1. Introduction 문제 제기 영상 합성 분야에서 Generative Adversarial Networks(GAN)는 큰 가능성을 보여주고 있지만, 고해상도 영상을 합성에는 큰 연산 비용과 많은 양의 학습 데이터가 요구되어 real-world 문제들에 적용하기 어렵다. 적은 양의 학습 데이터로 학습하기 위해 pre-training 후 fine-tuning 하는 기법이 제안되었으나, 1) 적합한 pre-training dataset을 찾기 어려울 수 있고, 2) pr...</summary> </entry> <entry><title>[Review] Differentiable Augmentation for Data-Efficient GAN Training</title><link href="https://donghweeyoon.github.io/posts/diffaugment-gan/" rel="alternate" type="text/html" title="[Review] Differentiable Augmentation for Data-Efficient GAN Training" /><published>2022-04-20T15:43:28+09:00</published> <updated>2022-04-20T15:43:28+09:00</updated> <id>https://donghweeyoon.github.io/posts/diffaugment-gan/</id> <content type="text/html" src="https://donghweeyoon.github.io/posts/diffaugment-gan/" /> <author> <name>Donghwee Yoon</name> </author> <category term="AI" /> <category term="Paper Review" /> <summary>Differentiable Augmentation for Data-Efficient GAN Training, Shengyu Zhao, et al., NeurIPS, 2020 본 논문은 GAN 학습에 적합한 data augmentation 기법을 제안한다. 같은 주제로 연구된 아래 논문을 함께 읽는 것을 추천한다. Training Generative Adversarial Networks with Limited Data, Tero Karras et al., NeurIPS, 2020 1. Problems GAN(Generative Adversarial Networks)은 영상 합성 기술을 크게 발전시켰다. 하지만 GAN은 high fidelity 영상을 합성하기 위해 많은 양의 데이터와 연산...</summary> </entry> </feed>
