https://zenodo.org/badge/866003411.svg

SynthVal#

SynthVal is a Python package developed to validate and verify the quality of synthetically generated data by comparing it to original data. The project focuses primarily on medical images, such as chest x-rays and mammographies, offering tools to compute similarity measures between original and synthetic datasets.

Purpose#

With the growing use of synthetic data in fields like healthcare and AI, it is essential to have reliable methods to evaluate how closely synthetic data resembles real data. SynthVal addresses this need by providing a straightforward framework for comparing original and synthetic data, enabling users to assess the quality and fidelity of synthetic datasets.

Key Features#

SynthVal is built around two main modules:

  1. Feature Extraction: The features_extraction.py module extracts vectors of features from images, capturing their essential characteristics to serve as the basis for similarity comparison.

  2. Similarity Metrics: The metrics.py module provides the capabilities to calculates several metrics to determine the similarity between original and synthetic datasets.

Table of Contents#