Verified project record

pytorch/vision

PyTorch Vision provides the core building blocks—datasets, transforms, and model architectures—needed for computer vision tasks in PyTorch. It is a comprehensive library offering popular public datasets, common image transformations, and pre-trained model architectures.

Project overview

As a computer vision companion to PyTorch, vision bundles dataset utilities, image transforms, and pre-trained model architectures into a single library, letting developers and researchers move from data loading to model inference without assembling separate tooling.

Project type: Model Development · Model Runtime · Image & Vision
Deployment: Refer to project documentation
License: BSD-3-Clause

Best for

Developers who need a programmatic library for loading datasets, applying image transforms, and using model architectures within PyTorch workflows.
AI engineers and researchers building or fine-tuning computer vision models who want pre-trained architectures and standard transforms in a single package.

Key capabilities

Downloads and prepares public datasets for machine learning, providing a standardized interface for data acquisition and preparation.
Provides common image transformations for computer vision, supporting both torch tensors and PIL images.
Supplies common and pre-trained model architectures for computer vision tasks.
Supports torch tensors and PIL images via Pillow or Pillow-SIMD for image processing.

Limitations and risks

Users are responsible for determining whether they have permission to use downloaded datasets under each dataset's license.
Pre-trained models may have their own licenses or terms derived from the training dataset, so users should verify usage rights before deployment.
SWAG models are released under the non-commercial CC-BY-NC 4.0 license, restricting them to non-commercial use.

Getting started

Install PyTorch (torch) before installing vision, because specific versions of PyTorch are required based on the version compatibility matrix.
After PyTorch is installed, install torchvision using the chosen installation method.

Evidence and sources

README: The torchvision package consists of popular datasets, model architectures, and common image transformations for computer vision.
README: This is a utility library that downloads and prepares public datasets. We do not host or distribute these datasets, vouch for their quality or fairness, or claim that you have lic…
README: title = {TorchVision: PyTorch's Computer Vision library}, author = {TorchVision maintainers and contributors}, year = 2016
Release: TorchVision 0.28.0 Release
README: The pre-trained models provided in this library may have their own licenses or terms and conditions derived from the dataset used for training.

Reviewed at: 2026-07-17T03:53:45Z