Geo-SAM

Geo SAM

🎉 GeoSAM is now on the QGIS Official Plugin Repository! Install and update Geo-SAM directly from the QGIS Plugin Manager -- Plugins > Manage and Install Plugins, search for "Geo SAM", and click Install. No manual download needed.

By Zhuoyi Zhao (Joey) and Chengyan Fan (Fancy) from Cryosphere Lab, ESSC, CUHK.

• Geo SAM
  • Introduction
  • What's New in GeoSAM v2.0
  • Installation and Usage
  • Demos
  • Citation
  • Acknowledgement

Introduction

Geo-SAM is a QGIS plugin that helps you segment, delineate, and label landforms efficiently in large geospatial raster images. It is built on the Segment Anything Model (SAM) and uses the strategies of encoding image features in advance and stripping the heavy image encoder from inference. The interactive segmentation runs at millisecond speeds on a laptop CPU, making it a convenient and efficient tool for remote sensing image analysis.

Since v2.0, Geo-SAM provides two segmentation workflows:

• Live Encoding mode -- select a raster layer and a SAM model, then start segmenting. Features are encoded on the fly using QGIS background tasks and cached for fast re-queries. No pre-encoding required.
• Pre-encoded mode -- use the Image Encoder to generate reusable feature files, then load them in the Segmentation tool. Best for repeated segmentation of the same image.

Comparison of the workflow between Geo-SAM and the original SAM. The original SAM encodes prompts and image simultaneously, while Geo-SAM encodes image features in advance and queries prompts at millisecond speeds by loading saved features.

What's New in GeoSAM v2.0

• Live Encoding mode -- no pre-encoding needed for quick exploration.
• geosam core library backbone -- replaces the internal implementation with a dedicated geospatial SAM library.
• SAM3 model support -- SAM, SAM 2, SAM 2.1, and SAM 3 model families.
• Plugin-managed dependencies -- one-click dependency installation into an isolated, plugin-private directory. No manual pip commands.
• Unified Settings dialog -- Dependencies, Model Management, Cache, and Help in one place.
• Online tile layer support -- segment directly from XYZ/WMS tile layers.
• Vectorization modes -- Pixel-Level or Simplified polygon output.
• Max Polygon Only mode -- keep only the largest polygon from each mask.
• Split-panel UI -- cleaner layout with Input/Output + Prompts on the left, Styles + Options on the right.
• Async preview pipeline -- smoother preview as you move the mouse.
• Multi-language UI -- the interface is translated into 中文, 日本語, 한국어, Français, Русский, العربية, Deutsch, Español, and Português, following the QGIS locale setting automatically.

See the changelog for the full list of changes.

Installation and Usage

Read the documentation for more details: https://geo-sam.readthedocs.io/en/latest/.

Quick start:

1. Install the Geo-SAM plugin from the QGIS Plugin Manager.
2. Open Geo-SAM Settings > Dependencies > Install Missing.
3. Restart QGIS.
4. Open Geo-SAM Settings > Model Management > download a model (e.g., SAM2.1 Base).
5. Click the Geo-SAM Segmentation icon, select an image layer and model, and start labelling.

See the Quick Start guide for a 5-minute walkthrough.

Demos

• Live Encoding Demo (millisecond-level response time after providing prompts)

• Image Encoder Demo (Old version, still available in the "Pre-encoded mode" since Geo-SAM v2.0)

• Encoder Copilot Demo

Citation

Zhao, Zhuoyi, Fan, Chengyan, & Liu, Lin. (2023). Geo SAM: A QGIS plugin using Segment Anything Model (SAM) to accelerate geospatial image segmentation (1.0). Zenodo. https://doi.org/10.5281/zenodo.8191039

@software{zhao_zhuoyi_2023_8191039,
  author       = {Zhao, Zhuoyi and Fan, Chengyan and Liu, Lin},
  title        = {{Geo SAM: A QGIS plugin using Segment Anything Model (SAM) to accelerate geospatial image segmentation}},
  year         = 2023,
  publisher    = {Zenodo},
  version      = {1.0},
  doi          = {10.5281/zenodo.8191039},
  url          = {https://doi.org/10.5281/zenodo.8191039}
}

Acknowledgement

This repo benefits from Segment Anything, SAM 2, Ultralytics, geosam, and QGIS. Thanks for their wonderful work.