Denoising Diffusion Probabilistic Models (DDPM)

Conditioning

Mechanism allowing to guide the generation process by providing additional information to the model, such as text, an image, or a class, often integrated via embeddings.

← Back