Conf. CVPR2024 Briefly review.
Reading objective : more enhancing personalization of optimize-based methods (e.g., Textual Inversion, DreamBooth) # Intro. 여러 personalization (ID preserve method)는 다양한 환경, 포즈, 객체 질감 등에 편향되어 있다.
이들을 적절히 해제함으로써 ID를 유지하면서 원하는 이미지를 생성할 수 있는데, 저자들은 VLM을 활용한 description을 추가하여 disentanglement를 효과적으로 수행한다. SID can be readily integrated with any optimization-based model, such as DreamBooth [45], Custom Diffusion [24], or SVDiff [15].
For the automated generation of SID, we...
#
AI
#
ID_Preservation
#
ImageEditing
#
LargeLanguageModels
#
MachineLearning
#
ObjectRemoval
#
Personalization
#
Prompt
#
SID
#
SVDiff
#
Text
#
GPT
#
GenerativeAI
#
ComputerVision
#
ControlledGeneration
#
CustomDiffusion
#
Customization
#
CVPR
#
CVPR2024
#
DeepLearning
#
Description
#
Disentanglement
#
DreamBooth
#
VLM
![Auto Encoder[오토인코더]](https://mblogthumb-phinf.pstatic.net/MjAyMjA5MDlfNzgg/MDAxNjYyNjYwMzQ3Nzcz.3OmDTTdniqebwD1UOH8vTwfmnhjj9iOtwtCGI77hiCAg.UWCR3VjQP4EUMGS8TCraWIuQqlybqfDnuK9DFxDdwTIg.PNG.jjunsss/image.png?type=w2)
