Amatis

Segmentation Benchmarking open-source software

A PyTorch research codebase for comparing two-stage semantic segmentation approaches under a shared protocol.

OpenGitHub repository

The repository contains model training, k-fold evaluation, inference, camera and ONNX export tools. It distributes no dataset: third-party inputs must be obtained from their original sources, and the restricted-access Amatis package is governed separately.

Scope
Nine architectures, twelve encoders and thirty-seven experiment configurations
Licence
MIT licence for source code only
Data
No dataset is distributed by the repository

What the public repository includes

The public repository provides source code, configuration examples and reproducibility documentation. It does not distribute wound-image datasets.