Skip to main content

A drop-in replacement for Torchvision Transforms using OpenCV

Project description


This repository is intended as a faster drop-in replacement for Pytorch's Torchvision augmentations. This repo uses OpenCV for fast image augmentation for PyTorch computer vision pipelines. I wrote this code because the Pillow-based Torchvision transforms was starving my GPU due to slow image augmentation.


  • A working installation of OpenCV. Tested with OpenCV version 3.4.1, 4.1.0
  • Tested on Windows 10 and Ubuntu 18.04. There is evidence that OpenCV doesn't work well with multithreading on Linux / MacOS, for example num_workers >0 in a pytorch DataLoader. I haven't run into this issue yet.


opencv_transforms is now a pip package! Simply use

  • pip install opencv_transforms


Breaking change! Please note the import syntax!

  • from opencv_transforms import transforms
  • From here, almost everything should work exactly as the original transforms.

Example: Image resizing

import numpy as np
image = np.random.randint(low=0, high=255, size=(1024, 2048, 3))
resize = transforms.Resize(size=(256,256))
image = resize(image)

Should be 1.5 to 10 times faster than PIL. See benchmarks


  • Most transformations are between 1.5X and ~4X faster in OpenCV. Large image resizes are up to 10 times faster in OpenCV.
  • To reproduce the following benchmarks, download the Cityscapes dataset.
  • An example benchmarking file can be found in the notebook bencharming_v2.ipynb I wrapped the Cityscapes default directories with a HDF5 file for even faster reading.

resize random crop change brightness change brightness and contrast change contrast only random horizontal flips

The changes start to add up when you compose multiple transformations together. composed transformations


  • <input type="checkbox" checked="" disabled="" /> Initial commit with all currently implemented torchvision transforms
  • <input type="checkbox" checked="" disabled="" /> Cityscapes benchmarks
  • <input type="checkbox" disabled="" /> Make the resample flag on RandomRotation, RandomAffine actually do something
  • <input type="checkbox" disabled="" /> Speed up augmentation in saturation and hue. Currently, fastest way is to convert to a PIL image, perform same augmentation as Torchvision, then convert back to np.ndarray

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for opencv-transforms, version 0.0.3.post0
Filename, size File type Python version Upload date Hashes
Filename, size opencv_transforms-0.0.3.post0.tar.gz (917.3 kB) File type Source Python version None Upload date Hashes View
Filename, size opencv_transforms-0.0.3.post0-py3-none-any.whl (18.5 kB) File type Wheel Python version py3 Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Huawei Huawei PSF Sponsor Microsoft Microsoft PSF Sponsor NVIDIA NVIDIA PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page