aleju/imgaug
Imgaug
imgaug is a Python library for machine learning data augmentation and computer vision dataset expansion. It provides tools to increase the volume and variety of training sets by applying random geometric, color, and noise transformations to images.
The library ensures spatial consistency by synchronizing transformations across images and their associated annotations, such as bounding boxes, keypoints, and segmentation maps. It uses a compositional pipeline pattern to chain multiple augmentations into sequences and employs deterministic seed management to reproduce specific data samples.
The project includes capabilities for parallel image processing to distribute workloads across multiple CPU cores. Additionally, it provides utilities for image metadata manipulation and tools for the visual inspection and rendering of augmented results.
Features
- Computer Vision Dataset Expansion - Increases training set size and diversity by generating stochastic variations of input images.
- Spatial Coordinate Synchronization - Ensures bounding box and keypoint coordinates are updated in synchronization with geometric image warping.
- Image Augmentation - Applies random geometric, color, and noise transformations to increase visual training variety.
- Augmentation - Synchronizes the application of identical random transformations across images and their corresponding annotations.
- Synthetic Dataset Generators - Creates modified versions of images and labels to expand the volume and variety of training datasets.
- Machine Learning Data Augmentation - Generates diverse variations of training images and labels to improve machine learning model robustness.
- Augmentation - Uses probability distributions to sample transformation values for diverse image variations.
- Synchronized Label Augmentations - Applying identical spatial transformations to images and their corresponding bounding boxes, keypoints, or segmentation maps.
- Annotation Alignment - Applies geometric transformations to bounding boxes and segmentation maps to maintain spatial accuracy.
- Coordinate System Mapping - Maps augmentation effects across diverse formats like segmentation maps and keypoints using a unified coordinate system.
- Deterministic - Provides controllable random seeds to ensure the exact reproduction of augmented data samples.
- Parallel Image Transformers - Utilizes parallel image transformers to distribute heavy processing tasks across multiple CPU cores.
- Compositional Transformation Pipelines - Implements a compositional pipeline pattern to chain multiple image augmentations into a single complex transformation.
- Distributed Task Queues - Distributes image processing workloads across multiple CPU cores using background processes.
- Computer Vision - Image augmentation library for machine learning.
- Computer Vision and Image Processing - Image augmentation library for machine learning.
- Data Wrangling - Library for image augmentation and keypoint transformation.
