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 |
Tags
- CVPR2022
- 베이지안 정리
- CIL
- Continual Learning
- DualPrompt
- Markov transition matrix
- Energy-based model
- mmcv
- Discrete diffusion
- L2P
- learning to prompt
- VQ-VAE
- requires_grad
- learning to prompt for continual learning
- VQ-diffusion
- timm
- ENERGY-BASED MODELS FOR CONTINUAL LEARNING
- Class Incremental Learning
- Face Alignment
- Class Incremental
- Vector Quantized Diffusion Model for Text-to-Image Synthesis
- Img2pose
- Mask diffusion
- Face Pose Estimation
- Mask-and-replace diffusion strategy
- PnP algorithm
- Facial Landmark Localization
- img2pose: Face Alignment and Detection via 6DoF
- prompt learning
- state_dict()
Archives
- Today
- Total
Computer Vision , AI
[One-page summary] TextTo 4D Dynamic Scene Generation by Singer et al. 본문
Paper_review[short]
[One-page summary] TextTo 4D Dynamic Scene Generation by Singer et al.
Elune001 2024. 1. 16. 00:29● Summary: Zero-shot 4D generation(time + 3D) using text prompt
● Approach highlight
-
HexPlane: represents a 4D scene with six planes of feature vectors spanning all pairs of axes in {X,Y,Z,T}.
-
To train Scene optimization, Project the rendered result, and denoising with text embedding and use this loss to train Scene optimization
● Main Results
● Discussion
- Limitations of representing complex or dynamic behaviors
'Paper_review[short]' 카테고리의 다른 글
'Paper_review[short]' Related Articles
more
