OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

X Unet

X Unet is a Python library implementing a configurable U-Net neural network in PyTorch, with nested U-Net stages, optional learned cross-stage attention residuals, and support shown for 2D images and 3D video or medical-image volumes.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

X Unet — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use X Unet when you need a configurable PyTorch U-Net implementation and want to experiment with nested stages or learned cross-stage attention, including for 3D-shaped inputs.

Who it is for

  • Python developers building or experimenting with U-Net-based image models.
  • Researchers and engineers exploring learned cross-stage attention or nested U-Net configurations.
  • Developers prototyping models for video or volumetric medical-image inputs.

Documented uses

  • Construct a 2D U-Net model and pass image-shaped tensors through it.
  • Configure nested stages and consolidated upsampling feature maps.
  • Replace standard skip connections with learned cross-stage attention residuals.
  • Set up a 3D model for video or CT/MRI-shaped tensors.
  • Use XAttnRes separately to route features across multiple scales.
Before you choose

Requirements & limits

The public repository provides installation instructions using pip and specifies an MIT license.

  • The examples show model construction and tensor input/output shapes rather than a complete training or evaluation pipeline.
  • Cross-stage attention can be enabled alongside skip connections or configured to replace them.
  • The repository includes 3D video and CT/MRI-shaped examples, while listing 3D memory-efficiency work as a todo.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install X Unet and construct a 2D model

Install the library and create the documented 2D XUnet model with a sample tensor.

Before you start

  • Python environment with PyTorch installed
  • Package installation via pip

Steps

  1. Install the package with the command below.

    Documented command
    pip install x-unet

  2. Import torch and XUnet from x_unet.

  3. Construct XUnet with dim=64, channels=3, dim_mults=(1, 2, 4, 8), nested_unet_depths=(7, 4, 2, 1), consolidate_upsample_fmaps=True, and use_xattnres=True.

  4. Create a sample tensor with torch.randn(1, 3, 256, 256) and pass it to the model.

Expected outcome

The example produces an output tensor with shape (1, 3, 256, 256).

Guide limitations

  • The example demonstrates model construction and a tensor pass; it does not document training, dataset preparation, or evaluation.
Documentation 1 ↗
Source-based practical guide

Construct the documented 3D X Unet model

Set up the repository's example model for video or CT/MRI-shaped inputs.

Before you start

  • Python environment with PyTorch installed
  • Package installed with pip install x-unet

Steps

  1. Import torch and XUnet from x_unet.

  2. Construct XUnet with dim=64, frame_kernel_size=3, channels=3, dim_mults=(1, 2, 4, 8), nested_unet_depths=(5, 4, 2, 1), consolidate_upsample_fmaps=True, weight_standardize=True, and use_xattnres=True.

  3. Create a sample tensor shaped (1, 3, 10, 128, 128) and pass it to the model.

Expected outcome

The example specifies an output with the same shape as its 3D input.

Guide limitations

  • The repository lists memory-efficiency improvements for 3D as todo items; the example does not establish performance or memory requirements.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
8 October 2026
Original website title
lucidrains/x-unet · Official homepage title · retained in its source language
How we review sourcesSuggest a correction