Skip to content

Welcome to ROXAS AI

ROXAS AI is an open-source, Python-based tool designed to streamline quantitative wood anatomy (QWA) and dendrochronology research. Built as an extensible plugin for the napari multi-dimensional image viewer, it enables researchers to efficiently extract anatomical traits and establish high-resolution time series of cell measurements within dated tree rings.

Currently optimized for linear conifer anatomies (with linear angiosperm anatomies currently in development), ROXAS AI provides an end-to-end workflow—from automated image segmentation to visual cross-dating—on Windows, macOS, and Linux.

Why ROXAS AI?

ROXAS AI was developed to modernize and expand upon the foundations of the classical ROXAS implementation. It addresses key limitations of classical workflows to make wood anatomy more accessible, automated, and scalable:

  • Cross-Platform Accessibility

    Runs natively on Windows, macOS, and Linux, bringing quantitative wood anatomy to all operating systems.

  • Open Science & Powerful Viewing

    Built on the open-source napari image viewer, completely eliminating dependencies on proprietary commercial software (such as Image-Pro Plus).

  • Modern AI Segmentation

    Leverages specialized deep-learning neural networks to segment cell lumens and ring boundaries with high accuracy and minimal manual editing.

  • Intuitive End-to-End Workflow

    Combines batch preparation, interactive raster/vector curation, visual cross-dating, and trait extraction into a streamlined, user-friendly interface.

High-Level Workflow

ROXAS AI guides you through an end-to-end pipeline from raw wood cross-section images to structured anatomical datasets:

flowchart TD
    A["<b>1. Prepare Images</b><br/>Standardize files & metadata"] --> B["<b>2. AI Segmentation</b><br/>Detect cells & ring boundaries"]
    B --> C["<b>3. Interactive Curation</b><br/>Review & cross-date rings"]
    C --> D["<b>4. Trait Measurement</b><br/>Export CSV tables & plots"]

Documentation Overview

Explore the sections below to get started:

  • Getting Started

    Installation instructions via Conda or pip, hardware acceleration setup, and how to launch the plugin.

  • User Guide

    Detailed walkthrough of the workflow widgets, editing tools, visual cross-dating, and settings.

  • Tutorials

    Step-by-step guides covering single-sample analysis and high-throughput batch processing pipelines.

  • Reference

    Technical specifications, output variables, deep-learning models, and shortcuts.

Scientific Background & Citation

If you use ROXAS AI in your research, please cite the following publications:

  1. Katzenmaier, M., Peters, R. L., & von Arx, G. (2025). Deep learning for automated tree-ring boundary detection in linear conifer wood anatomy. Frontiers in Plant Science, 16:1516635. doi:10.3389/fpls.2025.1516635

  2. Katzenmaier, M., Peters, R. L., & von Arx, G. (2023). Linear conifer cell segmentation using deep convolutional neural networks. Dendrochronologia, 81:126126. doi:10.1016/j.dendro.2023.126126

  3. von Arx, G., & Carrer, M. (2014). ROXAS - A new tool to build datasets of many thousand computerized wood-anatomical features per sample. Dendrochronologia, 32(3), 290–293. doi:10.1016/j.dendro.2013.12.001

Support & Contributing

ROXAS AI is actively developed and maintained by the Dendrosciences Group at the Swiss Federal Research Institute WSL.