StomataPy Documentation

StomataPy is a collection of resources to encourage community contributions for generalized stomatal segmentation.

1. We released generalized stomatal segmentation models trained on 400K diverse annotations

2. Community can correct minor model prediction errors with our annotation tool ISAT-SAM

3. The corrected annotations iteratively improve models, benefiting the entire community

Features

• Instance segmentation on stomatal complex; and pavment cells (require high quality images)

• Sementic segmentation of sub stomatal classes (stoma, outer ledge, pore)

• Derive cell area, length, width, rotation

• Customizable for other target such as trichome

• Batch analyze images and save results as Excel/JSON files

Quick Start

1. Installation: Clone the repository and install dependencies

2. Models: Download pre-trained models from our Hugging Face repository

3. Annotation: Use ISAT-SAM to create or correct annotations

4. Community: Contribute your annotations to improve models for everyone

Resources

• Hugging Face Models

• Dataset

• ISAT-SAM Tool

Note

This project is under active development. Feedback is welcome!

Contents

Documentation:

• Installation
  • Use Case 1: Google Colab (No Local GPU)
  • Use Case 2: Windows with NVIDIA GPU
  • Use Case 3: Linux Server with NVIDIA GPU