Denoising Diffusion Probabilistic Models (DDPM)

Resampling

Inference technique where multiple denoising trajectories are explored in parallel to improve the quality and diversity of generated samples.

← Indietro