🌳 GeoSAM2 Tree Segmentation

Developed by Mohammadreza Narimani & Sarbani Kumar
Digital Agriculture Laboratory | University of California, Davis

Deep Learning for Orchard Management

This project applies the Geospatial SAM2 model to automatically segment individual trees from high-resolution aerial imagery (Google & NAIP). By delineating each tree canopy and estimating canopy area, the tool supports orchard inventory, yield estimation, vigor analysis, and precision agriculture workflows.

Developed at the Digital Agriculture Laboratory, UC Davis as part of the UC Davis ESEARCH Program (Fall 2024), this tool leverages cutting-edge deep learning for agricultural applications.

✔️ SAM2 deep learning • Bounding-box assisted segmentation • Per-tree area estimation • Multi-orchard support

How the Model Works

1️⃣ Image Input

Upload high-resolution aerial imagery from NAIP, Google, or other sources for tree detection analysis.

2️⃣ SAM2 Processing

The Geospatial SAM2 model automatically identifies and segments individual tree canopies using deep learning.

3️⃣ Bounding Box Refinement

Bounding-box assisted segmentation improves accuracy by focusing on individual tree regions.

4️⃣ Output Generation

Generate per-tree polygons with canopy area estimates for orchard management and analysis.

Before & After Tree Segmentation

Example showing raw aerial imagery and GeoSAM2 tree detection results with automatic canopy delineation.

Mohammadreza Narimani GeoSAM2 tree segmentation deep learning results aerial imagery orchard management precision agriculture UC Davis Digital Agriculture Laboratory

GeoSAM2 Tree Segmentation Results

Try the Interactive App

Explore tree segmentation directly in your browser using the deep learning visualizer

Technical Information

Deep Learning Framework

  • Geospatial SAM2 (Segment Anything Model 2)
  • Segment-Geospatial framework by Dr. Qiusheng Wu
  • Bounding-box assisted segmentation
  • High-resolution aerial imagery support

Framework: Segment-Geospatial SAM2

Orchard Applications

  • Almond, walnut, pistachio orchards
  • Citrus and olive groves
  • Orchard inventory and tree counting
  • Yield estimation and vigor analysis
  • Precision agriculture workflows

Project Team

Development: Mohammadreza Narimani (PhD Candidate) & Sarbani Kumar (Undergraduate Mentee)

Program: UC Davis ESEARCH Fall 2024

Acknowledgment: Built on the Segment-Geospatial ecosystem by Dr. Qiusheng Wu and collaborators

Resources & Links

About the Developers

Mohammadreza Narimani (PhD Candidate) & Sarbani Kumar (Undergraduate Mentee)
University of California, Davis

Deep Learning • Remote Sensing • Digital Agriculture • Tree Segmentation • Precision Agriculture

Developing AI-powered tools for orchard management and agricultural automation at the Digital Agriculture Laboratory at UC Davis.

📧 mnarimani@ucdavis.edu | srbkumar@ucdavis.edu
🎓 Google Scholar Profile