Home

Exacta documentation

Exacta provides reference-validated segmentation metrics and differentiable losses for CPU and GPU. Its current focus is 2D and 3D medical image segmentation, including MRI anatomy and lesions. Definitions, empty-mask behavior and aggregation are explicit so results can be reproduced.

These pages document Exacta 0.0.1a5, including partially annotated evaluation, physical area/volume measurements, independent probability channels and class-distance training losses.

Start with the pages that match your task:

Task	Guide
Install Exacta	Installation
Run a complete example	Quickstart
Segment 2D medical images	2D medical segmentation
Evaluate hard segmentations	Metrics
Train with segmentation losses	Losses
Train/evaluate boundary, topology and probability objectives	Advanced segmentation
Match another tool's definition	Conventions
Evaluate separate lesions	Lesion evaluation
Choose and benchmark a backend	Performance
Work on the library	Development

The PyPI project, source repository, and issue tracker are maintained by Rodolfo Viana. Exacta is licensed under Apache 2.0. This is an alpha library; benchmark and validate the conventions relevant to your experiment before changing an established evaluation pipeline.