Fast, flexible, and advanced augmentation library for deep learning, computer vision, and medical imaging. Albumentations offers a wide range of transformations for both 2D (images, masks, bboxes, keypoints) and 3D (volumes, volumetric masks, keypoints) data, with optimized performance and seamless integration into ML workflows.
Project description
Albumentations
๐ฃ Stay updated! Subscribe to our newsletter for the latest releases, tutorials, and tips directly from the Albumentations team.
Docs | Discord | Twitter | LinkedIn
Albumentations is a Python library for image augmentation. Image augmentation is used in deep learning and computer vision tasks to increase the quality of trained models. The purpose of image augmentation is to create new training samples from the existing data.
Here is an example of how you can apply some pixel-level augmentations from Albumentations to create new images from the original one:
Why Albumentations
- Complete Computer Vision Support: Works with all major CV tasks including classification, segmentation (semantic & instance), object detection, and pose estimation.
- Simple, Unified API: One consistent interface for all data types - RGB/grayscale/multispectral images, masks, bounding boxes, and keypoints.
- Rich Augmentation Library: 70+ high-quality augmentations to enhance your training data.
- Fast: Consistently benchmarked as the fastest augmentation library also shown below section, with optimizations for production use.
- Deep Learning Integration: Works with PyTorch, TensorFlow, and other frameworks. Part of the PyTorch ecosystem.
- Created by Experts: Built by developers with deep experience in computer vision and machine learning competitions.
Community-Driven Project, Supported By
Albumentations thrives on developer contributions. We appreciate our sponsors who help sustain the project's infrastructure.
๐ Exclusive Partner |
---|
Your company could be here |
๐ก Integration Partner |
---|
Your company could be here |
๐ข Community Sponsor |
---|
๐ Become a Sponsor
Your sponsorship is a way to say "thank you" to the maintainers and contributors who spend their free time building and maintaining Albumentations. Sponsors are featured on our website and README. View sponsorship tiers on our support page
Table of contents
- Albumentations
- Why Albumentations
- Community-Driven Project, Supported By
- Table of contents
- Authors
- Installation
- Documentation
- A simple example
- Getting started
- Who is using Albumentations
- List of augmentations
- A few more examples of augmentations
- Benchmarking results
- Performance Comparison
- Contributing
- Community
- Citing
Authors
Current Maintainer
Vladimir I. Iglovikov | Kaggle Grandmaster
Emeritus Core Team Members
Mikhail Druzhinin | Kaggle Expert
Alexander Buslaev | Kaggle Master
Eugene Khvedchenya | Kaggle Grandmaster
Installation
Albumentations requires Python 3.9 or higher. To install the latest version from PyPI:
pip install -U albumentations
Other installation options are described in the documentation.
Documentation
The full documentation is available at https://albumentations.ai/docs/.
A simple example
import albumentations as A
import cv2
# Declare an augmentation pipeline
transform = A.Compose([
A.RandomCrop(width=256, height=256),
A.HorizontalFlip(p=0.5),
A.RandomBrightnessContrast(p=0.2),
])
# Read an image with OpenCV and convert it to the RGB colorspace
image = cv2.imread("image.jpg")
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
# Augment an image
transformed = transform(image=image)
transformed_image = transformed["image"]
Getting started
I am new to image augmentation
Please start with the introduction articles about why image augmentation is important and how it helps to build better models.
I want to use Albumentations for the specific task such as classification or segmentation
If you want to use Albumentations for a specific task such as classification, segmentation, or object detection, refer to the set of articles that has an in-depth description of this task. We also have a list of examples on applying Albumentations for different use cases.
I want to know how to use Albumentations with deep learning frameworks
We have examples of using Albumentations along with PyTorch and TensorFlow.
I want to explore augmentations and see Albumentations in action
Check the online demo of the library. With it, you can apply augmentations to different images and see the result. Also, we have a list of all available augmentations and their targets.
Who is using Albumentations
See also
List of augmentations
Pixel-level transforms
Pixel-level transforms will change just an input image and will leave any additional targets such as masks, bounding boxes, and keypoints unchanged. For volumetric data (volumes and 3D masks), these transforms are applied independently to each slice along the Z-axis (depth dimension), maintaining consistency across the volume. The list of pixel-level transforms:
- AdditiveNoise
- AdvancedBlur
- AutoContrast
- Blur
- CLAHE
- ChannelDropout
- ChannelShuffle
- ChromaticAberration
- ColorJitter
- Defocus
- Downscale
- Emboss
- Equalize
- FDA
- FancyPCA
- FromFloat
- GaussNoise
- GaussianBlur
- GlassBlur
- HEStain
- HistogramMatching
- HueSaturationValue
- ISONoise
- Illumination
- ImageCompression
- InvertImg
- MedianBlur
- MotionBlur
- MultiplicativeNoise
- Normalize
- PixelDistributionAdaptation
- PlanckianJitter
- PlasmaBrightnessContrast
- PlasmaShadow
- Posterize
- RGBShift
- RandomBrightnessContrast
- RandomFog
- RandomGamma
- RandomGravel
- RandomRain
- RandomShadow
- RandomSnow
- RandomSunFlare
- RandomToneCurve
- RingingOvershoot
- SaltAndPepper
- Sharpen
- ShotNoise
- Solarize
- Spatter
- Superpixels
- TextImage
- ToFloat
- ToGray
- ToRGB
- ToSepia
- UnsharpMask
- ZoomBlur
Spatial-level transforms
Spatial-level transforms will simultaneously change both an input image as well as additional targets such as masks, bounding boxes, and keypoints. For volumetric data (volumes and 3D masks), these transforms are applied independently to each slice along the Z-axis (depth dimension), maintaining consistency across the volume. The following table shows which additional targets are supported by each transform:
- Volume: 3D array of shape (D, H, W) or (D, H, W, C) where D is depth, H is height, W is width, and C is number of channels (optional)
- Mask3D: Binary or multi-class 3D mask of shape (D, H, W) where each slice represents segmentation for the corresponding volume slice
Transform | Image | Mask | BBoxes | Keypoints | Volume | Mask3D |
---|---|---|---|---|---|---|
Affine | โ | โ | โ | โ | โ | โ |
AtLeastOneBBoxRandomCrop | โ | โ | โ | โ | โ | โ |
BBoxSafeRandomCrop | โ | โ | โ | โ | โ | โ |
CenterCrop | โ | โ | โ | โ | โ | โ |
CoarseDropout | โ | โ | โ | โ | โ | โ |
ConstrainedCoarseDropout | โ | โ | โ | โ | โ | โ |
Crop | โ | โ | โ | โ | โ | โ |
CropAndPad | โ | โ | โ | โ | โ | โ |
CropNonEmptyMaskIfExists | โ | โ | โ | โ | โ | โ |
D4 | โ | โ | โ | โ | โ | โ |
ElasticTransform | โ | โ | โ | โ | โ | โ |
Erasing | โ | โ | โ | โ | โ | โ |
FrequencyMasking | โ | โ | โ | โ | โ | โ |
GridDistortion | โ | โ | โ | โ | โ | โ |
GridDropout | โ | โ | โ | โ | โ | โ |
GridElasticDeform | โ | โ | โ | โ | โ | โ |
HorizontalFlip | โ | โ | โ | โ | โ | โ |
Lambda | โ | โ | โ | โ | โ | โ |
LongestMaxSize | โ | โ | โ | โ | โ | โ |
MaskDropout | โ | โ | โ | โ | โ | โ |
Morphological | โ | โ | โ | โ | โ | โ |
Mosaic | โ | โ | โ | โ | ||
NoOp | โ | โ | โ | โ | โ | โ |
OpticalDistortion | โ | โ | โ | โ | โ | โ |
OverlayElements | โ | โ | ||||
Pad | โ | โ | โ | โ | โ | โ |
PadIfNeeded | โ | โ | โ | โ | โ | โ |
Perspective | โ | โ | โ | โ | โ | โ |
PiecewiseAffine | โ | โ | โ | โ | โ | โ |
PixelDropout | โ | โ | โ | โ | โ | โ |
RandomCrop | โ | โ | โ | โ | โ | โ |
RandomCropFromBorders | โ | โ | โ | โ | โ | โ |
RandomCropNearBBox | โ | โ | โ | โ | โ | โ |
RandomGridShuffle | โ | โ | โ | โ | โ | โ |
RandomResizedCrop | โ | โ | โ | โ | โ | โ |
RandomRotate90 | โ | โ | โ | โ | โ | โ |
RandomScale | โ | โ | โ | โ | โ | โ |
RandomSizedBBoxSafeCrop | โ | โ | โ | โ | โ | โ |
RandomSizedCrop | โ | โ | โ | โ | โ | โ |
Resize | โ | โ | โ | โ | โ | โ |
Rotate | โ | โ | โ | โ | โ | โ |
SafeRotate | โ | โ | โ | โ | โ | โ |
ShiftScaleRotate | โ | โ | โ | โ | โ | โ |
SmallestMaxSize | โ | โ | โ | โ | โ | โ |
SquareSymmetry | โ | โ | โ | โ | โ | โ |
ThinPlateSpline | โ | โ | โ | โ | โ | โ |
TimeMasking | โ | โ | โ | โ | โ | โ |
TimeReverse | โ | โ | โ | โ | โ | โ |
Transpose | โ | โ | โ | โ | โ | โ |
VerticalFlip | โ | โ | โ | โ | โ | โ |
XYMasking | โ | โ | โ | โ | โ | โ |
3D transforms
3D transforms operate on volumetric data and can modify both the input volume and associated 3D mask.
Where:
- Volume: 3D array of shape (D, H, W) or (D, H, W, C) where D is depth, H is height, W is width, and C is number of channels (optional)
- Mask3D: Binary or multi-class 3D mask of shape (D, H, W) where each slice represents segmentation for the corresponding volume slice
Transform | Volume | Mask3D | Keypoints |
---|---|---|---|
CenterCrop3D | โ | โ | โ |
CoarseDropout3D | โ | โ | โ |
CubicSymmetry | โ | โ | โ |
Pad3D | โ | โ | โ |
PadIfNeeded3D | โ | โ | โ |
RandomCrop3D | โ | โ | โ |
A few more examples of augmentations
Semantic segmentation on the Inria dataset
Medical imaging
Object detection and semantic segmentation on the Mapillary Vistas dataset
Keypoints augmentation
Benchmark Results
Image Benchmark Results
System Information
- Platform: macOS-15.1-arm64-arm-64bit
- Processor: arm
- CPU Count: 16
- Python Version: 3.12.8
Benchmark Parameters
- Number of images: 2000
- Runs per transform: 5
- Max warmup iterations: 1000
Library Versions
- albumentations: 2.0.4
- augly: 1.0.0
- imgaug: 0.4.0
- kornia: 0.8.0
- torchvision: 0.20.1
Performance Comparison
Number shows how many uint8 images per second can be processed on one CPU thread. Larger is better. The Speedup column shows how many times faster Albumentations is compared to the fastest other library for each transform.
Transform | albumentations 2.0.4 |
augly 1.0.0 |
imgaug 0.4.0 |
kornia 0.8.0 |
torchvision 0.20.1 |
Speedup (Alb/fastest other) |
---|---|---|---|---|---|---|
Affine | 1445 ยฑ 9 | - | 1328 ยฑ 16 | 248 ยฑ 6 | 188 ยฑ 2 | 1.09x |
AutoContrast | 1657 ยฑ 13 | - | - | 541 ยฑ 8 | 344 ยฑ 1 | 3.06x |
Blur | 7657 ยฑ 114 | 386 ยฑ 4 | 5381 ยฑ 125 | 265 ยฑ 11 | - | 1.42x |
Brightness | 11985 ยฑ 455 | 2108 ยฑ 32 | 1076 ยฑ 32 | 1127 ยฑ 27 | 854 ยฑ 13 | 5.68x |
CLAHE | 647 ยฑ 4 | - | 555 ยฑ 14 | 165 ยฑ 3 | - | 1.17x |
CenterCrop128 | 119293 ยฑ 2164 | - | - | - | - | N/A |
ChannelDropout | 11534 ยฑ 306 | - | - | 2283 ยฑ 24 | - | 5.05x |
ChannelShuffle | 6772 ยฑ 109 | - | 1252 ยฑ 26 | 1328 ยฑ 44 | 4417 ยฑ 234 | 1.53x |
CoarseDropout | 18962 ยฑ 1346 | - | 1190 ยฑ 22 | - | - | 15.93x |
ColorJitter | 1020 ยฑ 91 | 418 ยฑ 5 | - | 104 ยฑ 4 | 87 ยฑ 1 | 2.44x |
Contrast | 12394 ยฑ 363 | 1379 ยฑ 25 | 717 ยฑ 5 | 1109 ยฑ 41 | 602 ยฑ 13 | 8.99x |
CornerIllumination | 484 ยฑ 7 | - | - | 452 ยฑ 3 | - | 1.07x |
Elastic | 374 ยฑ 2 | - | 395 ยฑ 14 | 1 ยฑ 0 | 3 ยฑ 0 | 0.95x |
Equalize | 1236 ยฑ 21 | - | 814 ยฑ 11 | 306 ยฑ 1 | 795 ยฑ 3 | 1.52x |
Erasing | 27451 ยฑ 2794 | - | - | 1210 ยฑ 27 | 3577 ยฑ 49 | 7.67x |
GaussianBlur | 2350 ยฑ 118 | 387 ยฑ 4 | 1460 ยฑ 23 | 254 ยฑ 5 | 127 ยฑ 4 | 1.61x |
GaussianIllumination | 720 ยฑ 7 | - | - | 436 ยฑ 13 | - | 1.65x |
GaussianNoise | 315 ยฑ 4 | - | 263 ยฑ 9 | 125 ยฑ 1 | - | 1.20x |
Grayscale | 32284 ยฑ 1130 | 6088 ยฑ 107 | 3100 ยฑ 24 | 1201 ยฑ 52 | 2600 ยฑ 23 | 5.30x |
HSV | 1197 ยฑ 23 | - | - | - | - | N/A |
HorizontalFlip | 14460 ยฑ 368 | 8808 ยฑ 1012 | 9599 ยฑ 495 | 1297 ยฑ 13 | 2486 ยฑ 107 | 1.51x |
Hue | 1944 ยฑ 64 | - | - | 150 ยฑ 1 | - | 12.98x |
Invert | 27665 ยฑ 3803 | - | 3682 ยฑ 79 | 2881 ยฑ 43 | 4244 ยฑ 30 | 6.52x |
JpegCompression | 1321 ยฑ 33 | 1202 ยฑ 19 | 687 ยฑ 26 | 120 ยฑ 1 | 889 ยฑ 7 | 1.10x |
LinearIllumination | 479 ยฑ 5 | - | - | 708 ยฑ 6 | - | 0.68x |
MedianBlur | 1229 ยฑ 9 | - | 1152 ยฑ 14 | 6 ยฑ 0 | - | 1.07x |
MotionBlur | 3521 ยฑ 25 | - | 928 ยฑ 37 | 159 ยฑ 1 | - | 3.79x |
Normalize | 1819 ยฑ 49 | - | - | 1251 ยฑ 14 | 1018 ยฑ 7 | 1.45x |
OpticalDistortion | 661 ยฑ 7 | - | - | 174 ยฑ 0 | - | 3.80x |
Pad | 48589 ยฑ 2059 | - | - | - | 4889 ยฑ 183 | 9.94x |
Perspective | 1206 ยฑ 3 | - | 908 ยฑ 8 | 154 ยฑ 3 | 147 ยฑ 5 | 1.33x |
PlankianJitter | 3221 ยฑ 63 | - | - | 2150 ยฑ 52 | - | 1.50x |
PlasmaBrightness | 168 ยฑ 2 | - | - | 85 ยฑ 1 | - | 1.98x |
PlasmaContrast | 145 ยฑ 3 | - | - | 84 ยฑ 0 | - | 1.71x |
PlasmaShadow | 183 ยฑ 5 | - | - | 216 ยฑ 5 | - | 0.85x |
Posterize | 12979 ยฑ 1121 | - | 3111 ยฑ 95 | 836 ยฑ 30 | 4247 ยฑ 26 | 3.06x |
RGBShift | 3391 ยฑ 104 | - | - | 896 ยฑ 9 | - | 3.79x |
Rain | 2043 ยฑ 115 | - | - | 1493 ยฑ 9 | - | 1.37x |
RandomCrop128 | 111859 ยฑ 1374 | 45395 ยฑ 934 | 21408 ยฑ 622 | 2946 ยฑ 42 | 31450 ยฑ 249 | 2.46x |
RandomGamma | 12444 ยฑ 753 | - | 3504 ยฑ 72 | 230 ยฑ 3 | - | 3.55x |
RandomResizedCrop | 4347 ยฑ 37 | - | - | 661 ยฑ 16 | 837 ยฑ 37 | 5.19x |
Resize | 3532 ยฑ 67 | 1083 ยฑ 21 | 2995 ยฑ 70 | 645 ยฑ 13 | 260 ยฑ 9 | 1.18x |
Rotate | 2912 ยฑ 68 | 1739 ยฑ 105 | 2574 ยฑ 10 | 256 ยฑ 2 | 258 ยฑ 4 | 1.13x |
SaltAndPepper | 629 ยฑ 6 | - | - | 480 ยฑ 12 | - | 1.31x |
Saturation | 1596 ยฑ 24 | - | 495 ยฑ 3 | 155 ยฑ 2 | - | 3.22x |
Sharpen | 2346 ยฑ 10 | - | 1101 ยฑ 30 | 201 ยฑ 2 | 220 ยฑ 3 | 2.13x |
Shear | 1299 ยฑ 11 | - | 1244 ยฑ 14 | 261 ยฑ 1 | - | 1.04x |
Snow | 611 ยฑ 9 | - | - | 143 ยฑ 1 | - | 4.28x |
Solarize | 11756 ยฑ 481 | - | 3843 ยฑ 80 | 263 ยฑ 6 | 1032 ยฑ 14 | 3.06x |
ThinPlateSpline | 82 ยฑ 1 | - | - | 58 ยฑ 0 | - | 1.41x |
VerticalFlip | 32386 ยฑ 936 | 16830 ยฑ 1653 | 19935 ยฑ 1708 | 2872 ยฑ 37 | 4696 ยฑ 161 | 1.62x |
Contributing
To create a pull request to the repository, follow the documentation at CONTRIBUTING.md
Community
Citing
If you find this library useful for your research, please consider citing Albumentations: Fast and Flexible Image Augmentations:
@Article{info11020125,
AUTHOR = {Buslaev, Alexander and Iglovikov, Vladimir I. and Khvedchenya, Eugene and Parinov, Alex and Druzhinin, Mikhail and Kalinin, Alexandr A.},
TITLE = {Albumentations: Fast and Flexible Image Augmentations},
JOURNAL = {Information},
VOLUME = {11},
YEAR = {2020},
NUMBER = {2},
ARTICLE-NUMBER = {125},
URL = {https://www.mdpi.com/2078-2489/11/2/125},
ISSN = {2078-2489},
DOI = {10.3390/info11020125}
}
๐ซ Stay Connected
Never miss updates, tutorials, and tips from the Albumentations team! Subscribe to our newsletter.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
File details
Details for the file albumentations-2.0.8.tar.gz
.
File metadata
- Download URL: albumentations-2.0.8.tar.gz
- Upload date:
- Size: 354.5 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.1.0 CPython/3.9.22
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 |
4da95e658e490de3c34af8fcdffed09e36aa8a4edd06ca9f9e7e3ea0b0b16856
|
|
MD5 |
6d7f25f1868850fc401c746a222fece8
|
|
BLAKE2b-256 |
f4f485eb56c3217b53bcfc2d12e840a0b18ca60902086321cafa5a730f9c0470
|
File details
Details for the file albumentations-2.0.8-py3-none-any.whl
.
File metadata
- Download URL: albumentations-2.0.8-py3-none-any.whl
- Upload date:
- Size: 369.4 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.1.0 CPython/3.9.22
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 |
c4c4259aaf04a7386ad85c7fdcb73c6c7146ca3057446b745cc035805acb1017
|
|
MD5 |
950bd8cc13241b04c659a7e4602a1960
|
|
BLAKE2b-256 |
8e64013409c451a44b61310fb757af4527f3de57fc98a00f40448de28b864290
|