OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Image & Video AI

CUT

CUT is a Python and PyTorch implementation of unpaired image-to-image translation using patchwise contrastive learning and adversarial learning.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

CUT — OpenViST project profile in Image & Video AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use CUT when you need a local, code-based workflow for training and testing an unpaired image-translation model rather than a hosted image-editing service.

Who it is for

  • Researchers experimenting with contrastive image-translation methods
  • Developers training generative models to translate images between visual domains

Documented uses

  • Train a translation model without paired source and target images.
  • Compare the CUT and FastCUT training modes.
  • Run a trained model on a dataset and save translated results as an HTML page.
  • Explore the documented single-image training extension.
Before you choose

Requirements & limits

  • The documented workflow uses Python 3 and PyTorch 1.1 with additional dependencies.
  • The README lists Linux or macOS and a CPU or NVIDIA GPU with CUDA and cuDNN as prerequisites.
  • FastCUT is described by the project as using about half the GPU memory and training about twice as fast as CUT's cited comparison; actual results may vary.

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

Train and test the example grumpifycat model

Use the documented repository workflow to train CUT or FastCUT on the included example dataset and save test results.

Before you start

  • Linux or macOS
  • Python 3
  • A CPU or NVIDIA GPU with CUDA and cuDNN
  • PyTorch 1.1 and the listed dependencies, installed through the requirements file or Conda environment

Steps

  1. Clone the repository and install its Python dependencies.

  2. Download and unpack the grumpifycat dataset with the repository's dataset script.

  3. Run the documented training command with CUT_mode set to CUT or FastCUT.

  4. Run the documented test command for the trained model.

Expected outcome

Training checkpoints are saved under the checkpoints directory, and test results are written as an HTML file under results/grumpifycat/latest_train.

Guide limitations

  • The documented example uses the grumpifycat dataset; other datasets require their own data and training configuration.
  • The stated prerequisites and documented workflow do not guarantee performance or compatibility on a particular machine.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
18 October 2026
Original website title
taesungp/contrastive-unpaired-translation · Official homepage title · retained in its source language
How we review sourcesSuggest a correction