반응형 논문 정리42 Adding Conditional Control to Text-to-Image Diffusion Models (ControlNet) https://arxiv.org/abs/2302.05543 Adding Conditional Control to Text-to-Image Diffusion Models We present ControlNet, a neural network architecture to add spatial conditioning controls to large, pretrained text-to-image diffusion models. ControlNet locks the production-ready large diffusion models, and reuses their deep and robust encoding layers pr arxiv.org # Introduction Text-to-Image diffusio.. 2023. 11. 2. Benchmarking Robustness of Adaptation Methods onPre-trained Vision-Language Models (NeurIPS 2023) https://arxiv.org/abs/2306.02080 Benchmarking Robustness of Adaptation Methods on Pre-trained Vision-Language Models Various adaptation methods, such as LoRA, prompts, and adapters, have been proposed to enhance the performance of pre-trained vision-language models in specific domains. The robustness of these adaptation methods against distribution shifts have not been s arxiv.org Abstract : 본 논.. 2023. 10. 10. DiffusionEngine: Diffusion Model is Scalable Data Engine for Object Detection https://arxiv.org/abs/2309.03893 DiffusionEngine: Diffusion Model is Scalable Data Engine for Object Detection Data is the cornerstone of deep learning. This paper reveals that the recently developed Diffusion Model is a scalable data engine for object detection. Existing methods for scaling up detection-oriented data often require manual collection or generative m arxiv.org Abstract : object de.. 2023. 10. 9. Diffusion Models Beat GANs on Image Synthesis (NeurIPS 2021) https://arxiv.org/abs/2105.05233 Diffusion Models Beat GANs on Image Synthesis We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations. For conditional imag arxiv.org 참고 영상 : https://www.youtube.com/watch?v=gN1FQhQsUTE .. 2023. 5. 17. Zero-Shot Contrastive Loss for Text-Guided Diffusion Image Style Transfer https://arxiv.org/pdf/2303.08622.pdf Contribution - Zero-shot style transfer model 제안함 - 이를 위한 zero-shot contrastive loss 를 제안함 Overview of Proposed Method DDIM 으로 Forward 하고, Reverse 과정에서 DDPM을 사용. 그리고 Reverse 중간중간에 CLIP loss와 ZeCon loss를 추가해주면서 Style과 Content 에 대한 학습을 이룸. main contribution 중 하나는 추가적은 training 없이, guide 가 가능하다는 것이다. 이 때, CLIP으로 Style 을 guide 하고, ZeCon으로 content를 guide 함 Style G.. 2023. 4. 26. The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization https://arxiv.org/pdf/2006.16241.pdf Abstract 4개의 real-world distribution을 가진 dataset을 공개 : change image style , image blurriness, geographic location, camera operation, etc. 이러한 새로운 데이터셋으로 robustness를 위한 견고성을 개선하기 위한 방법을 테스트. 그 결과 larger model을 사용할수록 artificial data augmentations이 real-world distribution에서 robustness 하다는 것을 발견하였다. 데이터 증강이 효과가 있다는 것에 아이디어를 얻어 1000 배 더 많은 데이터로 pretrained 된 모델을 능가.. 2023. 3. 21. 이전 1 2 3 4 ··· 7 다음 반응형
