Skip to content

SynthShapes

SynthShapes is a primitive yet powerful PyTorch-based toolbox designed to generate synthetic 3D shapes, tailored specifically for augmenting biomedical imaging datasets. It is built to address the challenges faced in machine learning pipelines for biomedical imaging, such as limited availability of annotated data, the need for diverse training datasets, and the necessity of domain-specific augmentations.

MODULE DESCRIPTION
blending

Tools for alpha-blending tensors.

filters

Kernel-based spatial filters/transformations.

hist

Modules to map intensity histograms to one another.

shapes

Augmentation layers for incorporating synthetic shapes into 3D images.

texturizing

Tools to apply textures to label maps.

utils

General-purpose utilities for comon operations.