[Report Review] SORA, OpenAI
OpenAI의 SORA 발표 이후, 간단히 공부한 내용입니다.
Dmini's DL
[Paper Reivew] Adding Conditional Control to Text-to-Image Diffusion Models
Large pre-trained T2I freeze한 뒤, 원래 모델의 encoder layer를 복사해 strong backbone으로 사용하는 ControlNet제안.
[Paper Reivew] AnyText: Multilingual Visual Text Generation and Editing
Diffusion-based multilingual visual text generation and editing model AnyText 제안.
[Paper Reivew] UDiffText: A Unified Framework for High-quality Text Synthesis in Arbitrary Images via Character-aware Diffusion Models
SD의 Text encoder를 저자들의 character level text encoder로 대체하고 Model의 Cross-attention block만 denoising score matching loss, local attention loss, scene text recognition loss로 fine-tune.
[Paper Reivew] On Manipulating Scene Text in the Wild with Diffusion Models
One-shot style adaptation, Text-recognition guidance를 이용한 Diffusion based scene text editing을 수행하는 DBEST 제안.
[Paper Reivew] Multi-Concept Customization of Text-to-Image Diffusion
새로운 (single- or multiple-) concept을 Cross-attention layer의 Key, Value matrix만 fine-tune하여 학습할 수 있는 CustomDiffusion을 제안합니다.
[Paper Reivew] TextCtrl: Diffusion-based Scene Text Editing with Prior Guidance Control
STE 에서 Style, glyph를 나누어 diffusion model의 guidance로 활용하는 TextCtrl을 제안한 논문입니다.
[Paper Reivew] Autoregressive Image Generation without Vector Quantization
Continuous-valued space로 AR 모델을 일반화하는 MAR 제안합니다.
[Paper Reivew] Infinity: Scaling Bitwise AutoRegressive Modeling for High-Resolution Image Synthesis
Flow matching의 comprehensive and self-contained reviewd 입니다.
[Paper Reivew] Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction
Flow matching의 comprehensive and self-contained reviewd 입니다.