Monthly update (February 2025) As of 2025-03-01 (Changes in brackets for the past 30 days) Money: - $375 (+$70) in donations - 3 (-1) sponsors --- Monthly active Users in the UI tool - 253 (+24) ---- Stats: - Contributors: 5 (+3) - 7.2 million downloads (+16.7%) https://lnkd.in/g7sWxZhW - Used by 32,376 (+3%) https://lnkd.in/gVvcNxgP - Moved up 36 places at PyPI Download Leaderboard https://pypilb.vercel.app/. -------- Community: - 14634 (+0%) stars on GitHub - 2480 (+0%) citations for a scientific paper https://lnkd.in/g3hFuHwC - 161 (+4%) followers on Twitter https://lnkd.in/gT6CZE2a - 460 (+9%) followers on LinkedIn https://lnkd.in/gtPQVWQ6 - 12k (+9%) active visitors on the website ---- #albumentations #monthlyupdate #deeplearning #python #computervision #imageaugmentation #dataaugmentation #opensource #machinelearning #ai #datascience #projectstats
Albumentations
软件开发
San Francisco，CA 467 位关注者
Supercharge your image augmentation with Albumentations: Fast, flexible, and easy to use!
关于我们
Welcome to the official LinkedIn page for Albumentations, the leading image augmentation library designed for computer vision tasks. With a focus on performance, ease of use, and versatility, Albumentations provides an extensive toolkit for enhancing your machine-learning models by diversifying training data through high-quality image transformations. Developed by a team of passionate AI enthusiasts and researchers, Albumentations is built with Python and offers seamless integration with popular machine learning frameworks like TensorFlow and PyTorch. Whether you're working on image classification, segmentation, object detection, or any other computer vision challenge, Albumentations accelerates your projects by making image augmentation simpler, faster, and more effective. Why Albumentations? - Performance: Optimized for speed, Albumentations ensures your data augmentation doesn't become a bottleneck in the training process. - Comprehensive: From basic transformations like flips and rotations to advanced effects like color adjustments and complex composition, Albumentations covers all your augmentation needs. - Flexible and Easy to Use: A simple yet powerful API allows for easy integration into your existing workflows, making sophisticated augmentation strategies accessible to everyone. - Community-Driven: At the heart of Albumentations is a vibrant community of developers and researchers. Contributions, feedback, and discussions are always welcome, driving the library towards constant improvement and innovation. - Whether you're a seasoned data scientist, a machine learning enthusiast, or someone just starting in computer vision, Albumentations is your go-to library for transforming images into a powerful asset for model training. Join our community, contribute, and let's push the boundaries of what's possible in computer vision together.
- 网站
-
https://albumentations.ai/
Albumentations的外部链接
- 所属行业
- 软件开发
- 规模
- 2-10 人
- 总部
- San Francisco，CA
- 类型
- 非营利机构
- 领域
- Image Augmentation、Data Augmentation Techniques、Computer Vision、Deep Learning、Deep Learning、Real-time Augmentation、Image Preprocessing、AI Model Performance Improvement、Efficient Image Transformation、Custom Augmentation Pipelines、Integration with PyTorch and TensorFlow、Image Classification, Segmentation, and Detection、Advanced Geometric Transforms、Color Space Manipulation、Color Space Manipulation、Image Filtering、Batch Processing for Images、Augmentation for Object Detection、Randomized and Deterministic Augmentations、Performance Optimization in Image Processing和Handling Multimodal Data and Annotations
地点
-
主要
获取路线
US，CA，San Francisco
Albumentations员工
动态
-
?? Albumentations 2.0.5 is out ?? ?? https://lnkd.in/gJ4sF4E4 - 2.5 times speed up in GausssianBlur by Vedant Vijay Dalimkar - Speedups from 10 to 5500 percent in 10 helper functions by Saurabh Misra - Added SqureSymmetry, that is an alias for D4, but with a user friendly name. recommended to use it and not combinations of flips, transpose and rotations by 90 degrees. Supports image, mask, bounding boxes and keypoints #computervision #deeplearning #python #datascience
-
Improve data augmentation with Albumentations! ?? Albumentations offer powerful image augmentation with 70+ transformations like rotation, color adjustments, and noise addition, helping to create diverse and robust datasets. Key features: ? Uses OpenCV, NumPy, and SIMD for fast, efficient image augmentation on large datasets. ? Supports pixel, spatial, and mixed augmentations. ? Seamless integration with PyTorch and TensorFlow. Learn more ?? https://ow.ly/XcE850UHzcA
-
-
Big thanks to Abhishek Thakur for becoming a sponsor of Albumentations. Every dollar counts!
-
-
?? Albumentations 2.0.4 is out ?? ?? https://lnkd.in/grvQ4hMR - Added HEStain augmentation. https://lnkd.in/g3BudheH - 1.5 times speed up of GaussNoise: https://lnkd.in/gBj7S85K by Vedant Vijay Dalimkar - Fix in docs by Malte Ebner #computervision #deeplearning #python #datascience
-
-
Got feedback from user that after update np.random.seed() and random.seed() did not enforce reproducibility of the pipeline. They did, till they did not. A few months back moves random seed to Compose as setting seed externally was messing up other parts of the training pipeline. From FAQ:
-
-
?? Updated Albumentations Performance Benchmark + Real-world Impact ?? I've expanded our benchmarking to include transforms from Kornia AI and torchvision, testing all single image transforms on CPU (one core per image, RGB uint8). ?? Key findings: - Median speedup vs other libraries: 4.1x - 46/48 transforms show better performance in Albumentations ? Kornia outperforms in 2 cases: PlasmaShadow and LinearIllumination This helps identify areas where we need to improve Albumentations. ?? ?? Real-world validation: The Lightly team recently shared their experience switching to Albumentations, achieving: - 2x throughput improvement - GPU utilization increase from 66% to 99% - 50% reduction in training time and costs Full details in their blog: https://lnkd.in/gTz-Q2Pb ?? Want to verify these results? Benchmark code: https://lnkd.in/gsQUw934 ?? Your results may vary based on hardware configuration ?? All three libraries (torchvision, Albumentations, Kornia AI) have similar APIs, making it easy to test on your setup ?? If you spot any issues with the benchmark methodology or see different results on your hardware, please share. I use these benchmarks to identify optimization opportunities in Albumentations. #computervision #deeplearning #optimization #opensource
-
-
Albumentations 2.0.3 is out ?? https://lnkd.in/gZHQRsqK - Extended the functionality of the?strict?parameter in Compose. - Bugfix in filtering of bounding boxes based on aspect ratio - Speedup In?SaltAndPepper - Speedup in?AutoContrast - Speedup in Illumination - Speedup in?ElasticTransform - Speedup in?RandomRain - Bugfix in passing int?np.arrayas labels in BboxParams #computervision #deeplearning #python #datascience
-
Monthly update (January 2025. As of 2025-02 (Changes in brackets for the past 30 days) Money: - $305 (+$30) in donations - 4 (+2) sponsors --- Monthly active Users in the UI tool - 229 (+33) ---- Stats: - Contributors: 2 (-7) - 6.1 million downloads (+17.3%) https://lnkd.in/g7sWxZhW - Used by 31,246 (+1.03%) https://lnkd.in/gVvcNxgP - Moved up 55 places at PyPI Download Leaderboard https://pypilb.vercel.app/. -------- Community: - 14537 (+0%) stars on GitHub - 2460 (+2.5%) citations for a scientific paper https://lnkd.in/g3hFuHwC - 154 (+5) followers on Twitter https://lnkd.in/gT6CZE2a - 421 (+12) followers on LinkedIn https://lnkd.in/gtPQVWQ6 - 11k (+0%) active visitors on the website ---- #albumentations #monthlyupdate #deeplearning #python #computervision #imageaugmentation #dataaugmentation #opensource #machinelearning #ai #datascience #projectstats
-