https://github.com/pytorch/accimage

high performance image loading and augmenting routines mimicking PIL.Image interface
https://github.com/pytorch/accimage

Last synced: 3 months ago
JSON representation

high performance image loading and augmenting routines mimicking PIL.Image interface

• Host: GitHub
• URL: https://github.com/pytorch/accimage
• Owner: pytorch
• License: bsd-2-clause
• Archived: true
• Created: 2016-12-14T18:21:48.000Z (about 8 years ago)
• Default Branch: master
• Last Pushed: 2021-08-16T06:33:10.000Z (over 3 years ago)
• Last Synced: 2024-09-23T05:03:35.811Z (3 months ago)
• Language: C
• Size: 702 KB
• Stars: 317
• Watchers: 22
• Forks: 41
• Open Issues: 9
• Metadata Files:
  • Readme: README.md
  • Changelog: CHANGELOG.md
  • Contributing: CONTRIBUTING.md
  • License: LICENSE
  • Code of conduct: CODE_OF_CONDUCT.md

Awesome Lists containing this project

• awesome-list - accimage - High performance image loading and augmenting routines mimicking PIL.Image interface. (Data Processing / Data Pre-processing & Loading)

README

        
# accimage

[![Build status](https://github.com/pytorch/accimage/actions/workflows/accimage.yml/badge.svg)](https://github.com/pytorch/accimage/actions/workflows/accimage.yml)

[![Anaconda badge](https://anaconda.org/conda-forge/accimage/badges/version.svg)](https://anaconda.org/conda-forge/accimage)

An accelerated Image loader and preprocessor leveraging [Intel

IPP](https://software.intel.com/en-us/intel-ipp).

accimage mimics the PIL API and can be used as a backend for

[`torchvision`](https://github.com/pytorch/vision).

Operations implemented:

- `Image.resize((width, height))`

- `Image.crop((left, upper, right, lower))`

- `Image.transpose(PIL.Image.FLIP_LEFT_RIGHT)`

Enable the torchvision accimage backend with:

```python

torchvision.set_image_backend('accimage')

```

## Installation

accimage is available on conda-forge, simply run the following to install

```

$ conda install -c conda-forge accimage

```