U-Net
A neural network architecture shaped like a 'U' with an encoder that downsamples and a decoder that upsamples, connected by skip connections, widely used in image segmentation and diffusion models.
Architecture
The encoder progressively reduces spatial resolution while increasing feature channels. The decoder progressively restores resolution. Skip connections from encoder to decoder preserve fine-grained spatial information.
Applications
Originally designed for medical image segmentation. Now the core architecture in diffusion models (Stable Diffusion), where it performs the denoising. Also used in image-to-image translation, super-resolution, and inpainting.