일 | 월 | 화 | 수 | 목 | 금 | 토 |
---|---|---|---|---|---|---|
1 | 2 | 3 | 4 | 5 | 6 | 7 |
8 | 9 | 10 | 11 | 12 | 13 | 14 |
15 | 16 | 17 | 18 | 19 | 20 | 21 |
22 | 23 | 24 | 25 | 26 | 27 | 28 |
29 | 30 |
- ts
- Three
- DB
- C++
- Express
- figma
- Git
- SOLID
- API
- sqlite
- python
- Linux
- ps
- vscode
- mongo
- postgresql
- html
- CV
- frontend
- nodejs
- UI
- CSS
- PRISMA
- opencv
- js
- ML
- react
- review
- PyTorch
- GAN
- Today
- Total
목록ML (11)
아카이브

2020 ECCV에 투고된 논문입니다.https://www.matthewtancik.com/nerf?utm_source=chatgpt.com NeRF: Neural Radiance FieldsA method for synthesizing novel views of complex scenes by optimizing an underlying continuous volumetric scene function using a sparse set of input views.www.matthewtancik.com1. Introduction 본 연구에서는 장면 합성에서의 가장 고질적인 문제를 장면의 5차원 표현을 통해 매개변수를 직접 최적화하는 방법으로 해결했습니다. 이를 통해 입력된 이미지들과 실제 생성한 랜더링 ..

2023 CVPR에 투고된 논문입니다. 본문과 코드는 아래 github페이지에서 확인할 수 있습니다.https://diffusionrig.github.io/ DiffusionRig: Learning Personalized Priors for Facial Appearance EditingAbstract We address the problem of learning person-specific facial priors from a small number (e.g., 20) of portrait photos of the same person. This enables us to edit this specific person's facial appearance, such as expression and light..

KAIST 성민혁 교수님 연구실 소속 Jil Koo, Seungwoo Yoo, Minh Hieu Nguyen의 논문입니다. 2023 ICCV에 투고되었습니다. 본문과 코드는 아래 github페이지에서 확인할 수 있습니다.https://salad3d.github.io/ SALADShuffle! AutoSDF SALAD (Ours) AutoSDF SALAD (Ours) Text-Guided Part Completion Shuffle! Input Mesh Input Gaussians Output Mesh Output Gaussians Comparison of Shape Generation DPM PVD LION Voxel-GAN Neural Wavelet SPAGHETTI Diff. of $\mathbf{z}..

UC Berkeley의 Jonathan HO, Ajay Jain, Pieter Abbeel의 논문입니다. 본문과 코드는 아래 github에서 확인할 수 있습니다.GitHub - hojonathanho/diffusion: Denoising Diffusion Probabilistic Models GitHub - hojonathanho/diffusion: Denoising Diffusion Probabilistic ModelsDenoising Diffusion Probabilistic Models. Contribute to hojonathanho/diffusion development by creating an account on GitHub.github.com Abstract Diffusion probabi..

StyleGAN은 style transfer를 기반으로 인물의 구체적인 특징(style)을 구분하여 부분적으로 합성할 수 있게 해줍니다. 따라서 StyleGAN의 구조에 있어서 얼마나 style을 잘 잡아내는지, 그 style이 다른 style의 변화에 영향을 주지 않고 독립적인지를 따져보아야 합니다.1. 개요구조 설명에 앞서 몇 가지 용어를 알아야 합니다.1.1. Entanglement / disentanglement 앞서 말했듯이 styleGAN에서 중요한 점은 어떤 style이 다른 style에게 영향을 미쳐서는 안된다는 것입니다.만약 각각의 style들이 서로에게 영향을 주는 상태라면 이를 engtangled라고 부릅니다.반대로 각각의 style들이 독립적으로 존재한다면 disentangled라고..

NVIDIA Tero Karras, Smuli Laine, Timo Aila의 논문입니다. 본문과 코드는 아래 github에서 확인할 수 있습니다.https://github.com/NVlabs/stylegan GitHub - NVlabs/stylegan: StyleGAN - Official TensorFlow ImplementationStyleGAN - Official TensorFlow Implementation. Contribute to NVlabs/stylegan development by creating an account on GitHub.github.com Abstract 본 논문은 style transfer literature에 기반한 새로운 GAN 구조를 고안했습니다. 이 새로운 구조는 사..

일반적인 GAN의 경우 생성하는 위조 데이터를 원하는 속성의 데이터로 생성해낼 수 없다는 단점이 있습니다. 예를 들어 강아지 사진을 생성하는 GAN 모델의 경우, 무작위 강아지 사진을 생성할 수 는 있지만, 푸들이나 리트리버와 같이 특정한 견종을 지정하여 생성할 수 없습니다. 이를 개선하기 위해 GAN에 추가적인 조건(condition)을 제공하여 구체적인 속성의 데이터를 생성할 수 있게 해주는 조건적 적대 생성 신경망(Conditional Generative Adversarial Network, CGAN)이 개발되었습니다. [1] 1. 구조 CGAN은 원하는 조건을 나타내는 condition 변수를 가지고 있음을 제외하면 GAN과 동일한 구조를 가지고 있습니다.이 condition 변수는 상황에 따라..

적대적 생성 신경망(Genrative Adversarial Neural network, GAN) 혹은 적대적 생성 모델은 신경망의 응용 기술 중 하나로, 주로 주어진 상황에 맞는 위조 데이터를 만들어내는 데 사용됩니다. 예를 들어 GAN은 가짜 인물사진이나 문장을 만들어낼 수 있습니다. GAN은 2014년 신경정보시스템처리학회에(NIPS)에서 Ian Goodfellow에 의해 처음으로 제시되었고, 이후로 여러 변형과 응용을 거치며 널리 쓰이게 되었습니다. 1. 개요 GAN은 크게 Generator(생성자)와 Discriminator(판별자) 두 개의 모델로 이루어져 있으며, 이들이 서로 적대적으로 경쟁하면서 각자의 성능을 향상시키게 됩니다. - Generator : 위조값을 생성합니다. Discrim..