-
25-Gaussian
The data distribution is taken to be a 25-dimensional Gaussian distribution, generated by mixture of gaussians. -
Simple SDE
The data distribution is taken to be a one-dimensional Gaussian distribution, P0(x0) = N (x0 | 0, v0). -
ImageNet and CC-3M datasets
ImageNet [9] and CC-3M [43] datasets -
LSUN-{cat, bedroom, church} and FFHQ datasets
LSUN-{cat, bedroom, church} [48] and FFHQ [25] datasets -
CelebA and Chairs
CelebA and Chairs are the two datasets utilised by Higgins et al. (2016) for qualitative evaluation. -
AutoStudio: Crafting Consistent Subjects in Multi-turn Interactive Image Gene...
AutoStudio is a multi-agent training-free framework that addresses multi-turn interactive image generation successfully. -
Face-related tasks with ControlNet
The dataset used in this paper is a face-related task with ControlNet. The dataset contains images with different poses and conditions. -
Style-guided generation with Stable Diffusion
The dataset used in this paper is a style-guided generation task with Stable Diffusion. The dataset contains images with different styles and prompts. -
Denoising Diffusion Implicit Models
The dataset used in the paper for training and testing the Denoising Diffusion Implicit Models. -
Diffusion Exponential Integrator Sampler
The dataset used in this paper is a diffusion model for image generation. -
ImageNet 64x64, ImageNet 128x128, LSUN 256x256
The dataset used in the paper is not explicitly described, but it is mentioned that the authors used pre-trained DDPMs on ImageNet 64x64, ImageNet 128x128, and LSUN 256x256. -
DDIM or PLMS
Text-to-image models, such as DDIM or PLMS. -
OpenCLIP and LAION-5B
Language-guided image diffusion models, such as OpenCLIP and LAION-5B. -
DALL-E3 and Stable Diffusion Dataset
A dataset used by the authors to test their hypothesis about the white bear phenomenon in large models. -
White Bear Phenomenon Dataset
A dataset generated by the authors to test their hypothesis about the white bear phenomenon in large models. -
CelebA, FFHQ, and Teapots
CelebA, FFHQ, and Teapots datasets for image generation. -
High-Fidelity Image Generation With Fewer Labels
High-fidelity image generation with fewer labels