algorave 2.1.1

Fast, flexible, and advanced augmentation library for deep learning, computer vision, and medical imaging. Algorave 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.

Algorave

A fast and flexible image augmentation library for deep learning, computer vision, and machine learning workflows.

Algorave is a fork of the Albumentations library, providing powerful image transformation capabilities with a focus on performance and ease of use.

Installation

pip install algorave

For development installation:

git clone https://github.com/your-username/algorave.git
cd algorave
pip install -e .

Features

• Fast and efficient: Optimized for performance with NumPy and OpenCV backends
• Flexible: Supports a wide range of image augmentations for various computer vision tasks
• Easy to use: Simple, intuitive API that integrates seamlessly with popular deep learning frameworks
• Extensible: Easy to add custom augmentations
• Battle-tested: Based on the proven Albumentations library used in numerous production systems

Quick Start

import algorave as A
import cv2

# Define 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
image = cv2.imread("image.jpg")
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# Apply augmentations
transformed = transform(image=image)
transformed_image = transformed["image"]

Supported Data Types

Algorave supports augmentation of:

• Images
• Masks
• Bounding boxes
• Keypoints

Requirements

• Python >= 3.9
• NumPy
• OpenCV
• PyYAML
• scikit-image (optional)

Contributing

Contributions are welcome! Please see CONTRIBUTING.md for guidelines.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

Algorave is based on Albumentations, originally created by the Albumentations team. This fork was created from commit 66212d7.

Special thanks to the original Albumentations authors:

• Vladimir Iglovikov
• Alexander Buslaev
• Alex Parinov
• Eugene Khvedchenya
• Mikhail Druzhinin