🖼️ DDPM Image Generator

Generates a new image from pure Gaussian noise using a Denoising Diffusion Probabilistic Model trained from scratch in PyTorch.

Click Generate to run the full reverse diffusion process. The right panel shows intermediate denoising steps so you can watch the image emerge from noise.

Model: Custom U-Net (64→128→256 channels) trained with MSE loss on image noise.
Assignment: Generative AI (AI4009) — Spring 2026, NUCES.