Notice
Recent Posts
Recent Comments
Link
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 | 31 |
Tags
- VirtualTryON
- sound-to-image generation
- diffusion
- 네이버 부스트캠프 ai tech 6기
- dreamfusion
- text-to-video diffusion
- posterior distillation sampling
- novel view synthesis
- 3d editting
- visiontransformer
- objectdrop
- DP
- dreamgaussian
- BOJ
- magic clothing
- transformer
- 3d editing
- sonicdiffusion
- 3d gaussian splatting
- 코테
- Programmers
- Vit
- 3d generation
- insturctnerf2nerf
- instructany2pix
- Python
- 논문리뷰
- 프로그래머스
- 코딩테스트
- text2room
Archives
- Today
- Total
목록ControlNet (1)
평범한 필기장
![](http://i1.daumcdn.net/thumb/C150x150/?fname=https://blog.kakaocdn.net/dn/cw6LnN/btsF7Hr0osV/LHDvZZ6Kz13dbTs1i1fHX1/img.png)
이번 논문은 최종 프로젝트에서 메인으로 쓰일 ControlNet이다.프로젝트에 관한 포스팅은 여기에서 확인할 수 있다.Why ControlNet?본 논문에서 ControlNet을 왜 제시했을까? 일단 먼저 기존의 text-to-image generation diffusion모델을 생각해보자. Stable Diffusion을 통해 우리는 고퀄리티의 이미지를 텍스트를 통해 생성할 수 있었다. 하지만 본 논문에서는 이러한 text-to-image generation에는 두 가지 한계점이 존재한다고 말하고 있다.Text만으로는 이미지의 spatial composition을 제공하는 것에는 한계가 있다.특정 condition에 대한 데이터셋이 작기 때문에 직접 finetuning하거나 continued..
AI/Diffusion
2024. 3. 27. 00:31