ORBiS TREE · ENGLISH DOCUMENT
Concepts →Diffusion Model
A generative model that learns to reverse a gradual noise process to create data such as images.
한국어
English
This English page is a curated translation layer linked to the Korean source article. Community changes are currently made on the Korean source, where the full revision history and anonymous edit trail are preserved.
Diffusion models are generative models that learn a reverse process from noise toward structured data. They became especially prominent in high-quality image generation, although related techniques apply beyond images.
How it works
Training typically teaches a model to predict noise or another quantity at different noise levels. At generation time, a sampler repeatedly transforms an initial noise signal toward a sample that matches the learned data distribution and conditioning information.
Why it matters
Diffusion models helped drive modern text-to-image systems and multimodal research. Their quality and speed depend on model architecture, conditioning, sampler design, number of steps, and training data.
Related concepts
SOURCES
Sources
-
Denoising Diffusion Probabilistic ModelsUC Berkeley / arXivOpen source ↗
KNOWLEDGE LINKS