SegMap

Table of Contents

• Introduction
• Getting Started
  • Installation
  • Setup
  • Basic Usage
  • Keyboard Shortcuts
• Advanced Topics
  • Self-Hosted API Server
  • Troubleshooting
• Project Information
  • Contributing
  • License
  • Authors and References

Introduction

SegMap is a QGIS plugin that uses deep learning to enable fast map digitization. It allows users to extract objects from aerial imagery with just a few clicks, replacing traditional labor-intensive polygon editing.

Benefits:

• 90% Efficiency Gain: Complete in seconds with 1-5 clicks what traditionally requires hundreds of vertex adjustments
• AI-Powered: Leverages deep learning models for precise object extraction
• QGIS Integration: Works seamlessly within your existing GIS workflow

SegMap revolutionizes tasks like land cover mapping, infrastructure monitoring, and environmental analysis by combining AI efficiency with intuitive user interaction.

Compatibility Note: SegMap is tested and supported on QGIS 3.40-Bratislava LTR. Other versions may not work as expected.

Getting Started

Installation

1. Download the latest release from the releases page
2. In QGIS: Plugins > Manage and Install Plugins... > Install from ZIP
3. Select the downloaded ZIP file and click Install Plugin

Setup

1. Click SegMap: Settings to configure server and token
   • Quick start: Use demo server https://segmap.nodes.studio with token demo
2. Click SegMap: Start to open the tool panel
3. Select your Model, Raster, Output, and edit the Class list as needed

Basic Usage

1. Click Start to begin
2. Navigate to your target object (middle mouse button/scroll to pan/zoom)
3. Left-click on the object, right-click on background areas to refine selection
4. Click Confirm to save to the vector layer
5. Click End when finished

Keyboard Shortcuts

• Enter: Start/Confirm segmentation
• Left Arrow: Undo last click
• Right Arrow: Redo click
• Esc: End segmentation

Advanced Topics

Self-Hosted API Server

For self-hosting instructions, see server/README.md.

Privacy Notice

• Images from your canvas are sent to the server for processing but not stored
• No user data is collected or retained after processing
• For sensitive data, use a self-hosted server in your secure environment

Troubleshooting

Common issues and solutions:

Issue	Solution
Plugin Loading Failure	• Verify QGIS 3.40-Bratislava LTR • Reinstall the plugin • Check QGIS log panel for errors
Server Connection Issues	• Check internet connection • Try again later (demo server may be down) • Consider self-hosting
Poor Results	• Add more clicks to refine selection • Use right-clicks to exclude unwanted areas • Center the object in view • Try different models
Performance Issues	• Self-host on a more powerful machine • Reduce image resolution • Close other resource-intensive applications

Project Information

Contributing

We welcome contributions! Here's how:

• Report Issues: Use GitHub issue tracker with detailed steps and screenshots
• Develop: Fork, branch, code, test, and submit pull requests
• Document: Improve docs, create translations, or develop tutorials

All contributions are licensed under GPL-3.0.

License

GPL-3.0 License. See LICENSE file for details.

Authors and References

Authors: Shoukun Sun, Zhe Wang

Citations:

@article{cfr_icl_2024,
      title={CFR-ICL: Cascade-Forward Refinement with Iterative Click Loss for Interactive Image Segmentation},
      volume={38},
      url={https://ojs.aaai.org/index.php/AAAI/article/view/28306},
      DOI={10.1609/aaai.v38i5.28306},
      number={5},
      journal={Proceedings of the AAAI Conference on Artificial Intelligence},
      author={Sun, Shoukun and Xian, Min and Xu, Fei and Capriotti, Luca and Yao, Tiankai},
      year={2024},
      month={Mar.},
      pages={5017-5024}
}

@misc{is_aerial_2024,
      title={Interactive segmentation in aerial images: a new benchmark and an open access web-based tool}, 
      author={Zhe Wang and Shoukun Sun and Xiang Que and Xiaogang Ma},
      year={2024},
      eprint={2308.13174},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2308.13174}, 
}