### Install cli-anything-samgeo

Source: https://github.com/opengeos/segment-geospatial/blob/main/agent-harness/cli_anything/samgeo/README.md

Install the command-line tool by navigating to the agent-harness directory and running the pip install command. This makes the `cli-anything-samgeo` command available in your system's PATH.

```bash
cd agent-harness
pip install -e .
```

--------------------------------

### Example: GeoJSON Mask Polygons

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/api.md

Use this example to get segmentation masks as GeoJSON polygons. Ensure the image file and prompt are correctly provided.

```bash
curl -X POST http://localhost:8000/segment/text \
  -F "file=@image.tif" \
  -F "prompt=building" \
  -F "output_format=geojson"
```

--------------------------------

### Install SamGeo with Pixi

Source: https://github.com/opengeos/segment-geospatial/blob/main/README.md

Instructions for installing SamGeo using pixi, a recommended package and environment manager. This includes creating a new pixi project, editing the configuration, and installing dependencies.

```bash
# Install pixi (Linux/macOS)
curl -fsSL https://pixi.sh/install.sh | sh

# Or on Windows (PowerShell)
powershell -ExecutionPolicy Bypass -c "irm -useb https://pixi.sh/install.ps1 | iex"

# Create a new pixi project
pixi init geo
cd geo

# Edit pixi.toml with your configuration (see docs for GPU/CPU examples)
# Then install
pixi install

# Start Jupyter Lab
pixi run jupyter lab
```

--------------------------------

### Install segment-geospatial

Source: https://github.com/opengeos/segment-geospatial/blob/main/agent-harness/cli_anything/samgeo/skills/SKILL.md

Install the segment-geospatial library with all extras and the agent-harness locally.

```bash
pip install segment-geospatial[all]
cd agent-harness && pip install -e .
```

--------------------------------

### Set up virtual environment and install dependencies

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/contributing.md

Set up a virtual environment using virtualenvwrapper and install the project in development mode.

```shell
$ mkvirtualenv segment-geospatial
$ cd segment-geospatial/
$ python setup.py develop
```

--------------------------------

### Install Pixi environment

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/installation.md

Installs all dependencies defined in the pixi.toml file for the current project.

```bash
pixi install
```

--------------------------------

### Claude Code Example: Download Tiles and Get Info

Source: https://github.com/opengeos/segment-geospatial/blob/main/agent-harness/cli_anything/samgeo/README.md

An example of Claude Code using cli-anything-samgeo to download OpenStreetMap tiles for a specified bounding box and then retrieve information about the downloaded image. This demonstrates chaining commands for data acquisition and inspection.

```bash
cli-anything-samgeo --json data download-tiles -o portland.tif -b "-122.68,45.51,-122.66,45.53" -z 17
cli-anything-samgeo --json data info portland.tif
```

--------------------------------

### Install Required Packages

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/image_captioning.ipynb

Install the necessary packages for geospatial analysis and image captioning. Uncomment the line to run the installation.

```python
# %pip install "segment-geospatial[samgeo3]"
```

--------------------------------

### Install Segment-Geospatial

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam2_video.ipynb

Install the required segment-geospatial library. Uncomment and run this cell to install.

```python
# %pip install -U segment-geospatial
```

--------------------------------

### Install necessary libraries

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/workshops/AIforGood_2025.ipynb

Installs the segment-geospatial and groundingdino-py libraries. Run this command in your environment before proceeding.

```python
%pip install segment-geospatial groundingdino-py
```

--------------------------------

### Install Pixi on Windows

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/installation.md

Installs Pixi using a PowerShell command. Close and re-open PowerShell after installation.

```powershell
powershell -ExecutionPolicy Bypass -c "irm -useb https://pixi.sh/install.ps1 | iex"
```

```powershell
pixi --version
```

--------------------------------

### Install QGIS Plugin using install_plugin.py

Source: https://github.com/opengeos/segment-geospatial/blob/main/qgis-samgeo-plugin/README.md

Steps to clone the plugin repository, activate the conda environment, and install the plugin using the provided Python script.

```bash
# Clone the repository
git clone https://github.com/opengeos/qgis-samgeo-plugin.git
cd qgis-samgeo-plugin
conda activate geo

# Install (Linux/macOS)
python install_plugin.py
```

--------------------------------

### Manual Plugin Installation via Symlink

Source: https://github.com/opengeos/segment-geospatial/blob/main/qgis-samgeo-plugin/README.md

Instructions for manually installing the plugin by creating a symbolic link to the plugin directory within the QGIS plugins folder.

```bash
ln -s /path/to/qgis-samgeo-plugin ~/.local/share/QGIS/QGIS3/profiles/default/python/plugins/samgeo_plugin
```

--------------------------------

### Install QGIS Plugin using install_plugin.sh

Source: https://github.com/opengeos/segment-geospatial/blob/main/qgis-samgeo-plugin/README.md

Alternative method to install the plugin using a shell script after cloning the repository and activating the conda environment.

```bash
# Clone the repository
git clone https://github.com/opengeos/qgis-samgeo-plugin.git
cd qgis-samgeo-plugin
conda activate geo

# Or use the shell script
chmod +x install_plugin.sh
./install_plugin.sh
```

--------------------------------

### Install API Dependencies

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/api.md

Install the necessary dependencies for the API with the 'api' extra. To include a specific SAM model backend, combine extras.

```bash
pip install "segment-geospatial[api]"

```

```bash
pip install "segment-geospatial[api,samgeo3]"

```

--------------------------------

### Start the API Server with Uvicorn

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/api.md

Alternatively, start the segment-geospatial API server directly using 'uvicorn'.

```bash
uvicorn samgeo.api:app --host 0.0.0.0 --port 8000

```

--------------------------------

### Install necessary libraries

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/workshops/IPPN_2024.ipynb

Installs the leafmap library with raster support, segment-geospatial, and d2spy. Also installs a specific version of numpy.

```python
# %pip install -U "leafmap[raster]" segment-geospatial d2spy
```

```python
# %pip install numpy==1.26.4
```

--------------------------------

### Download and display an example image

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_point_prompts.ipynb

Download a sample image from a URL and display it using the show_image function. This prepares the image for segmentation tasks.

```python
url = "https://raw.githubusercontent.com/facebookresearch/sam3/refs/heads/main/assets/images/truck.jpg"
image_path = download_file(url)
```

```python
show_image(image_path, axis="on")
```

--------------------------------

### Install Dependencies

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam2_automatic.ipynb

Install the required libraries for segment-geospatial and leafmap. Ensure you are using a GPU runtime for optimal performance.

```python
# %pip install -U segment-geospatial leafmap
```

--------------------------------

### Install Required Packages

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/detectree2.ipynb

Install PyTorch, Detectron2, detectree2, and segment-geospatial. Adjust PyTorch installation for your CUDA version.

```python
# Install PyTorch (adjust for your CUDA version)
# %pip install torch torchvision

# Install Detectron2
# %pip install 'git+https://github.com/facebookresearch/detectron2.git'

# Install detectree2
# %pip install git+https://github.com/PatBall1/detectree2.git

# Install segment-geospatial
# %pip install segment-geospatial
```

--------------------------------

### Verify PyTorch and CUDA installation

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/installation.md

Runs a Python script using Pixi to check PyTorch version, CUDA availability, and GPU name. Essential for confirming GPU setup.

```bash
pixi run python -c "import torch; print('PyTorch:', torch.__version__); print('CUDA available:', torch.cuda.is_available()); print('GPU:', (torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'None'))"
```

--------------------------------

### Install Dependencies

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_image_segmentation_jpg.ipynb

Installs the necessary libraries for SAM3 image segmentation. Ensure you have the correct version of transformers.

```python
# %pip install "segment-geospatial[samgeo3]"
```

```python
# %pip install transformers==5.0.0rc0
```

--------------------------------

### Install QGIS and SamGeo with Conda

Source: https://github.com/opengeos/segment-geospatial/blob/main/qgis-samgeo-plugin/README.md

Recommended installation procedure using conda to create a dedicated environment for QGIS and the segment-geospatial package.

```bash
conda create -n geo python=3.12
conda activate geo
conda install -c conda-forge qgis segment-geospatial
```

--------------------------------

### Example: JSON Output (Pixel-Coordinate Bounding Boxes)

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/api.md

This example demonstrates how to obtain segmentation results as JSON, with bounding boxes defined by pixel coordinates. Suitable for direct image analysis.

```bash
curl -X POST http://localhost:8000/segment/text \
  -F "file=@image.jpg" \
  -F "prompt=building" \
  -F "output_format=json"
```

--------------------------------

### Install segment-geospatial

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/box_prompts.ipynb

Install the required library for geospatial segmentation. This should be run in a GPU-enabled environment.

```python
# %pip install segment-geospatial
```

--------------------------------

### Install Pixi on Linux/macOS

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/installation.md

Installs Pixi using a curl command. Close and re-open your terminal or reload your shell after installation.

```bash
curl -fsSL https://pixi.sh/install.sh | sh
```

```bash
pixi --version
```

--------------------------------

### Start the API Server

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/api.md

Start the segment-geospatial API server using the 'samgeo-api' command. Options are available for custom host/port, preloading models, and auto-reloading during development.

```bash
samgeo-api

```

```bash
samgeo-api --host 0.0.0.0 --port 8000        # Custom host/port

```

```bash
samgeo-api --preload sam2:sam2-hiera-large     # Preload a model at startup

```

```bash
samgeo-api --reload                            # Auto-reload for development

```

--------------------------------

### Install Dependencies

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/fast_sam.ipynb

Install the necessary Python packages for segmenting geospatial data with FastSAM. Uncomment and run this cell to begin.

```python
# %pip install segment-geospatial segment-anything-fast
```

--------------------------------

### Install SAM 3 on Windows with Conda and Pip

Source: https://github.com/opengeos/segment-geospatial/blob/main/README.md

Specific commands for installing SAM 3 on Windows, including creating a Python 3.12 environment, installing PyTorch with CUDA support, segment-geospatial, and other necessary tools.

```bash
conda create -n geo python=3.12
conda activate geo
conda install pytorch torchvision pytorch-cuda=12.1 -c pytorch -c nvidia
pip install "segment-geospatial[samgeo3]"
pip install triton-windows ipykernel jupyterlab
```

--------------------------------

### Install Segment Geospatial from PyPI

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/installation.md

Use this command to install the library using pip. This is the standard method for most Python projects.

```bash
pip install segment-geospatial
```

--------------------------------

### Install SamGeo Dependencies with Pip

Source: https://github.com/opengeos/segment-geospatial/blob/main/qgis-samgeo-plugin/README.md

Installs all necessary SamGeo dependencies, including those only available on PyPI, using pip.

```bash
pip install -U "segment-geospatial[samgeo3]"
```

--------------------------------

### Start QGIS with SamGeo Plugin Environment

Source: https://github.com/opengeos/segment-geospatial/blob/main/qgis-samgeo-plugin/README.md

Commands to activate the conda environment and launch QGIS, making the SamGeo plugin available.

```bash
conda activate geo
qgis
```

--------------------------------

### Start Jupyter Lab

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/installation.md

Launches Jupyter Lab using Pixi, providing an environment to use segment-geospatial.

```bash
pixi run jupyter lab
```

--------------------------------

### Example: Detections Output (Geographic-Coordinate Bounding Boxes)

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/api.md

Use this example to retrieve segmentation results with bounding boxes in geographic coordinates. This format is useful for GIS applications.

```bash
curl -X POST http://localhost:8000/segment/text \
  -F "file=@image.tif" \
  -F "prompt=building" \
  -F "output_format=detections"
```

--------------------------------

### Run Segmentation with Box Prompts (All Similar Objects)

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/api.md

This example demonstrates segmenting an image using bounding box coordinates. For SAM3, this will find all similar objects within the image, not just those inside the box. Specify the image file and bounding box coordinates.

```bash
curl -X POST http://localhost:8000/segment/predict \
  -F "file=@image.tif" \
  -F "boxes=[[10, 20, 300, 400]]" \
  -F "output_format=geojson"
```

--------------------------------

### Install segment-geospatial with PyPI

Source: https://github.com/opengeos/segment-geospatial/blob/main/README.md

Install segment-geospatial with specific dependencies for SAMGeo 3. Other options include [samgeo], [samgeo2], [fast], [hq], [text], [fer], and [api].

```bash
pip install "segment-geospatial[samgeo3]" # Or any other choice of the above
```

--------------------------------

### Import required libraries

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/workshops/AIforGood_2025.ipynb

Imports leafmap for interactive mapping and SamGeo2 for segmentation. Ensure these libraries are installed.

```python
import leafmap
from samgeo import SamGeo2, regularize
```

--------------------------------

### Install Segment Geospatial from GitHub

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/installation.md

Install the development version directly from the GitHub repository using pip. This is useful for accessing the latest features or contributing to the project.

```bash
pip install git+https://github.com/opengeos/segment-geospatial
```

--------------------------------

### Generate Segmentation Mask and Convert to Vector

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/usage.md

This example demonstrates generating a segmentation mask from a satellite image and converting it to a GeoPackage vector file. Ensure you have the SAM checkpoint file downloaded.

```python
import os
import torch
from samgeo import SamGeo
from samgeo.common import tms_to_geotiff

bbox = [-95.3704, 29.6762, -95.368, 29.6775]
image = 'satellite.tif'
tms_to_geotiff(output=image, bbox=bbox, zoom=20, source='Satellite')

out_dir = os.path.join(os.path.expanduser('~'), 'Downloads')
checkpoint = os.path.join(out_dir, 'sam_vit_h_4b8939.pth')

device = 'cuda' if torch.cuda.is_available() else 'cpu'
sam = SamGeo(
    checkpoint=checkpoint,
    model_type='vit_h',
    device=device,
    erosion_kernel=(3, 3),
    mask_multiplier=255,
    sam_kwargs=None,
)

mask = 'segment.tif'
sam.generate(image, mask)

vector = 'segment.gpkg'
sam.tiff_to_gpkg(mask, vector, simplify_tolerance=None)
```

--------------------------------

### Install Dependencies

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam2_text_prompts.ipynb

Installs the necessary Python packages for geospatial image segmentation using SAM 2 and related libraries. Ensure you have a compatible environment.

```python
# %pip install segment-geospatial groundingdino-py leafmap localtileserver
```

--------------------------------

### Install optional Conda dependencies

Source: https://github.com/opengeos/segment-geospatial/blob/main/README.md

Install optional dependencies like groundingdino-py and segment-anything-fast for enhanced functionality within your conda environment.

```bash
conda install -c conda-forge groundingdino-py segment-anything-fast
```

--------------------------------

### Import Segment Geospatial Library

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/usage.md

Import the necessary library to start using segment-geospatial.

```python
import samgeo
```

--------------------------------

### Get Bounding Boxes in Pixel Coordinates

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/api.md

This example demonstrates how to get bounding boxes in pixel coordinates for non-georeferenced images. It sends a JPEG image and a prompt to the API, expecting a JSON output containing detection details.

```python
# Get bounding boxes in pixel coordinates (suitable for non-georeferenced images)
with open("image.jpg", "rb") as f:
    response = requests.post(
        url,
        files={"file": ("image.jpg", f, "image/jpeg")},
        data={"prompt": "car", "output_format": "json"},
    )

result = response.json()
for det in result["detections"]:
    print(f"Object {det['id']}: bbox={det['bbox']}, score={det['score']:.3f}")
```

--------------------------------

### Import Libraries and Download Video

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_video_masks.ipynb

Import necessary libraries and download a sample video file for demonstration.

```python
import os
import numpy as np
from samgeo import SamGeo3Video, download_file
```

```python
url = "https://github.com/opengeos/datasets/releases/download/videos/cars.mp4"
video_path = download_file(url)
```

--------------------------------

### List Available Models

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/api.md

Retrieve a list of available model versions and IDs, along with which models are currently loaded in memory, using the GET /models endpoint.

```bash
curl http://localhost:8000/models

```

--------------------------------

### Claude Code Example: Segment Buildings and Export

Source: https://github.com/opengeos/segment-geospatial/blob/main/agent-harness/cli_anything/samgeo/README.md

An example of how Claude Code might execute a sequence of cli-anything-samgeo commands to segment buildings and export the results as a GeoPackage. Note the use of `--json` for machine-readable output and the correct order of flags.

```bash
cli-anything-samgeo --json project new -n buildings -o project.json -s satellite.tif -t sam2
cli-anything-samgeo --json --project project.json segment automatic -o masks.tif
cli-anything-samgeo --json vector convert masks.tif buildings.gpkg
```

--------------------------------

### Install Optional Conda Dependencies

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/installation.md

Install optional dependencies like groundingdino-py and segment-anything-fast using mamba. These are not included in the default installation.

```bash
mamba install -c conda-forge groundingdino-py segment-anything-fast
```

--------------------------------

### Initialize ImageCaptioner

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/image_captioning.ipynb

Create an instance of the ImageCaptioner. Customize BLIP and spaCy models, or specify the device for inference.

```python
captioner = ImageCaptioner(
    blip_model_name="Salesforce/blip-image-captioning-base",
    spacy_model_name="en_core_web_sm",
)
```

--------------------------------

### Low-level mask prediction with box prompt

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_point_prompts.ipynb

Perform instance segmentation using a bounding box prompt via the `predict_inst` method. This example demonstrates generating a single mask for the specified box.

```python
# Box prompt with Python list
masks, scores, logits = sam.predict_inst(
    box=[425, 600, 700, 875],
    multimask_output=False,
)

sam.show_inst_masks(masks, scores, box_coords=[425, 600, 700, 875])
```

--------------------------------

### Create a new Pixi project

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/installation.md

Initializes a new Pixi project named 'geo' and navigates into its directory.

```bash
pixi init geo
cd geo
```

--------------------------------

### Set Up Model Checkpoint and Output Directory

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/satellite-predictor.ipynb

Specify the output directory for results and the local path to the SAM model checkpoint file.

```python
out_dir = os.path.join(os.path.expanduser("~"), "Downloads")
checkpoint = os.path.join(out_dir, "sam_vit_h_4b8939.pth")
```

--------------------------------

### Create a New Project

Source: https://github.com/opengeos/segment-geospatial/blob/main/agent-harness/cli_anything/samgeo/README.md

Initialize a new segmentation project. Specify a project name, output file for project configuration, and the input satellite image.

```bash
cli-anything-samgeo project new -n my-seg -o project.json -s image.tif
```

--------------------------------

### Run unit tests

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/contributing.md

Execute all unit tests for the project by navigating to the root directory and running the discover command.

```bash
python -m unittest discover tests/
```

--------------------------------

### Install segment-geospatial with conda

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/index.md

Installs segment-geospatial and its dependencies from the conda-forge channel into the active conda environment.

```bash
conda install -c conda-forge segment-geospatial
```

--------------------------------

### Install Segment Geospatial with Conda

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/installation.md

Recommended for creating a dedicated environment. This installs segment-geospatial and its dependencies using conda and mamba.

```bash
conda create -n geo python
conda activate geo
conda install -c conda-forge mamba
mamba install -c conda-forge segment-geospatial
```

--------------------------------

### Visualize Sample Image

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/detectree2.ipynb

Create a leafmap interactive map and add the downloaded sample tree image as a raster layer.

```python
m = leafmap.Map()
m.add_raster(image_path, layer_name="Tree Image")
m
```

--------------------------------

### Install segment-geospatial with Conda (GPU PyTorch)

Source: https://github.com/opengeos/segment-geospatial/blob/main/README.md

If your system has a GPU, use this command to force the installation of the GPU version of PyTorch along with segment-geospatial.

```bash
conda install -c conda-forge segment-geospatial "pytorch=*=cuda*"
```

--------------------------------

### Install segment-geospatial with Conda

Source: https://github.com/opengeos/segment-geospatial/blob/main/README.md

Create a new conda environment named 'geo', activate it, and install segment-geospatial from conda-forge. This is recommended for managing dependencies.

```bash
conda create -n geo python
conda activate geo
conda install -c conda-forge segment-geospatial
```

--------------------------------

### Enter Interactive REPL

Source: https://github.com/opengeos/segment-geospatial/blob/main/agent-harness/cli_anything/samgeo/README.md

Run the cli-anything-samgeo command without any arguments to enter an interactive Read-Eval-Print Loop (REPL) for command exploration.

```bash
cli-anything-samgeo
```

--------------------------------

### Initialize SamGeo Model

Source: https://github.com/opengeos/segment-geospatial/blob/main/qgis-samgeo-plugin/README.md

Initializes the SamGeo model with specified backend and configuration. Use 'cuda' for GPU acceleration if available.

```python
from samgeo import SamGeo3

# Initialize the model
sam = SamGeo3(
    backend="meta",
    enable_inst_interactivity=True,
    confidence_threshold=0.5,
)
```

--------------------------------

### Create and activate conda environment

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/index.md

Recommended for installing segment-geospatial with conda. This creates a new environment named 'geo', installs Python, and then activates the environment.

```bash
conda create -n geo python
```

```bash
conda activate geo
```

--------------------------------

### Download Sample Data

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam2_video.ipynb

Download sample remote sensing data in a zip archive. Ensure all images have the same width and height.

```python
url = "https://github.com/opengeos/datasets/releases/download/raster/landsat_ts.zip"
leafmap.download_file(url)
```

--------------------------------

### Initialize Leafmap and Add Raster

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_image_segmentation.ipynb

Initialize a leafmap interactive map and add the downloaded satellite image as a raster layer for visualization.

```python
m = leafmap.Map()
m.add_raster(image_path, layer_name="Satellite image")
m
```

--------------------------------

### Initialize Video Tracker

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_video_prompts.ipynb

Initialize the tracker for the video. This prepares the SamGeo3Video object for point-prompted segmentation and propagation.

```python
sam.init_tracker()
```

--------------------------------

### List Export Formats

Source: https://github.com/opengeos/segment-geospatial/blob/main/agent-harness/cli_anything/samgeo/skills/SKILL.md

List all available export formats supported by the tool.

```bash
cli-anything-samgeo export formats
```

--------------------------------

### Download Sample Data

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_tiled_segmentation.ipynb

Downloads a sample satellite image (GeoTIFF) for demonstration purposes. Replace `image_path` with your own large GeoTIFF.

```python
# Download a sample satellite image
url = "https://huggingface.co/datasets/giswqs/geospatial/resolve/main/naip/naip_water_train.tif"
image_path = "naip_water_train.tif"

if not os.path.exists(image_path):
    common.download_file(url, image_path)
```

--------------------------------

### Data Information

Source: https://github.com/opengeos/segment-geospatial/blob/main/agent-harness/cli_anything/samgeo/skills/SKILL.md

Get information about a raster data file.

```bash
cli-anything-samgeo data info RASTER_PATH
```

--------------------------------

### Download SAM 3.1 model

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/installation.md

Downloads the SAM 3.1 model from Hugging Face Hub using Pixi. Requires prior authentication.

```bash
pixi run hf download facebook/sam3.1
```

--------------------------------

### Prepare Training Data

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/detectree2.ipynb

Prepare image tiles and corresponding manual crown polygons for training a custom detectree2 model. Includes splitting into training and testing sets.

```python
# Prepare training and test data
# train_dir, test_dir = prepare_training_data(
#     image_path="path/to/orthomosaic.tif",
#     crowns_path="path/to/manual_crowns.gpkg",
#     output_dir="./training_data",
#     tile_width=40,
#     tile_height=40,
#     buffer=30,
#     threshold=0.6,  # Minimum crown coverage per tile
#     test_fraction=0.15,  # Fraction for testing
# )
```

--------------------------------

### Get Coordinate Reference System (CRS)

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/workshops/IPPN_2024.ipynb

Extract and print the Coordinate Reference System (CRS) of the raster data.

```python
crs = data.rio.crs.to_string()
print(crs)
```

--------------------------------

### Initialize SamGeo3

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_tiled_segmentation.ipynb

Initializes the SamGeo3 object with the 'meta' backend for processing.

```python
sam = SamGeo3(backend="meta")
```

--------------------------------

### Extract Features from Caption

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/image_captioning.ipynb

Use the extract_features() method to get noun features from an existing caption, with options for filtering.

```python
# Extract features from an existing caption
features = captioner.extract_features(caption)
print(f"All Features: {features}")

aerial_features = captioner.extract_features(caption, include_features="default")
print(f"Aerial Features: {aerial_features}")
```

--------------------------------

### Initialize LangSAM Model

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam2_text_prompts.ipynb

Initializes the `LangSAM` class with the 'sam2-hiera-large' model. This step may take a few minutes as it downloads model weights.

```python
sam = LangSAM(model_type="sam2-hiera-large")
```

--------------------------------

### Get D2S project boundary

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/workshops/IPPN_2024.ipynb

Retrieves the project boundary as a GeoJSON formatted Python dictionary for the selected D2S project.

```python
# Get project boundary as Python dictionary in GeoJSON structure
project_boundary = project.get_project_boundary()
project_boundary
```

--------------------------------

### Import necessary libraries

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/box_prompts.ipynb

Import the leafmap library for interactive mapping, tms_to_geotiff for downloading imagery, and SamGeo for segmentation.

```python
import leafmap
from samgeo.common import tms_to_geotiff
from samgeo import SamGeo
```

--------------------------------

### Initialize FastSAM Model

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/fast_sam.ipynb

Initialize the SamGeo class with a specified FastSAM model. This step may take a few minutes as it downloads model weights.

```python
from samgeo.fast_sam import SamGeo

sam = SamGeo(model="FastSAM-x.pt")
```

--------------------------------

### Define Text Prompt

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam2_text_prompts.ipynb

Sets the text prompt to guide the segmentation process. The model will identify objects corresponding to this prompt.

```python
text_prompt = "tree"
```

--------------------------------

### Run segment-geospatial with Docker

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/installation.md

Use this command to start a Docker container for segment-geospatial. It maps port 8888 for Jupyter Notebook access.

```bash
docker run -it -p 8888:8888 giswqs/segment-geospatial:latest
```

--------------------------------

### CLI Syntax

Source: https://github.com/opengeos/segment-geospatial/blob/main/agent-harness/cli_anything/samgeo/skills/SKILL.md

Basic syntax for the cli-anything-samgeo command, including global flags for JSON output and project paths.

```bash
cli-anything-samgeo [--json] [--project PATH] COMMAND [ARGS...]
```

--------------------------------

### Get data products from a flight

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/workshops/IPPN_2024.ipynb

Retrieves a collection of data products associated with a specific flight. Iterates through and prints each data product.

```python
# Get list of data products from a flight
data_products = flight.get_data_products()

for data_product in data_products:
    print(data_product)
```

--------------------------------

### Download Sample Image

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/detectree2.ipynb

Download a sample tree image from a specified URL for testing purposes.

```python
image_url = (
    "https://github.com/opengeos/datasets/releases/download/samgeo/tree_image.tif"
)
image_path = leafmap.download_file(image_url)
```

--------------------------------

### Initialize SamGeo2 Predictor for Video

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam2_video.ipynb

Initialize the SamGeo2 predictor with a specific model ID and enable video processing. Use a GPU runtime for better performance.

```python
predictor = SamGeo2(
    model_id="sam2-hiera-large",
    video=True,
)
```

--------------------------------

### Define Area of Interest

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/satellite-predictor.ipynb

Get the bounding box for the area of interest from user interaction on the map, or use a default bounding box.

```python
if m.user_roi_bounds() is not None:
    bbox = m.user_roi_bounds()
else:
    bbox = [-123.0127, 44.9957, -122.9874, 45.0045]
```

--------------------------------

### Health Check Endpoint

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/api.md

Check the server status and version using the GET /health endpoint. This is a basic check to ensure the API is running.

```bash
curl http://localhost:8000/health

```

```json
{"status": "ok", "version": "1.2.3"}

```

--------------------------------

### Run Segmentation with Point Prompts

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/api.md

Use this endpoint to segment an image using point coordinates and labels. Ensure the image file is uploaded correctly and specify the desired output format.

```bash
curl -X POST http://localhost:8000/segment/predict \
  -F "file=@image.tif" \
  -F "point_coords=[[100, 200]]" \
  -F "point_labels=[1]" \
  -F "output_format=geojson"
```

--------------------------------

### Get CHM statistics

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/workshops/IPPN_2024.ipynb

Calculates and returns statistics for a Cloud Optimized GeoTIFF (COG) layer, such as min, max, mean, and standard deviation.

```python
leafmap.cog_stats(chm_url_202206)
```

--------------------------------

### Get DSM statistics

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/workshops/IPPN_2024.ipynb

Calculates and returns statistics for a Cloud Optimized GeoTIFF (COG) layer, such as min, max, mean, and standard deviation.

```python
leafmap.cog_stats(dsm_url_202206)
```

--------------------------------

### Get Data Information with JSON Output

Source: https://github.com/opengeos/segment-geospatial/blob/main/agent-harness/cli_anything/samgeo/README.md

Obtain detailed information about a geospatial data file in JSON format. This is useful for programmatic access to metadata.

```bash
cli-anything-samgeo --json data info image.tif
```

--------------------------------

### Download and Load Video

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_video_segmentation.ipynb

Download a sample MP4 video file from a URL and set it as the input for the SamGeo3Video predictor.

```python
url = "https://github.com/opengeos/datasets/releases/download/videos/cars.mp4"
video_path = download_file(url)
```

```python
sam.set_video(video_path)
```

--------------------------------

### Display Map for Interactive Prompting

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/workshops/AIforGood_2025.ipynb

Displays the leafmap, allowing users to interactively draw bounding boxes or points on the map to guide the segmentation process.

```python
m
```

--------------------------------

### Filter data products by data type

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/workshops/IPPN_2024.ipynb

Filters a DataProductCollection to include only products of a specified data type. Example shows filtering for 'ortho' data products.

```python
# Example of creating new collection of data products with the "ortho" data type
ortho_data_products = data_products.filter_by_data_type("ortho")
print(ortho_data_products)
```

--------------------------------

### Initialize SamGeo Model

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/satellite.ipynb

Initialize the SamGeo model with a specified model type ('vit_h') and checkpoint file. The checkpoint file should be available in the same directory or a specified path.

```python
sam = SamGeo(
    model_type="vit_h",
    checkpoint="sam_vit_h_4b8939.pth",
    sam_kwargs=None,
)
```

--------------------------------

### Download and Load Video

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_object_tracking.ipynb

Download a sample video file from a URL and set it for processing using the SamGeo3Video object. Supports MP4, JPEG frame directories, and GeoTIFF directories.

```python
url = "https://huggingface.co/datasets/giswqs/geospatial/resolve/main/basketball.mp4"
video_path = download_file(url)
```

```python
sam.set_video(video_path)
```

```python
sam.show_video(video_path)
```

--------------------------------

### Define Area of Interest (ROI)

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam2_automatic.ipynb

Get the bounding box for the region of interest from user interaction on the map, or use a default bounding box if no ROI is drawn.

```python
if m.user_roi_bounds() is not None:
    bbox = m.user_roi_bounds()
else:
    bbox = [-95.3704, 29.6762, -95.368, 29.6775]
```

--------------------------------

### Import Libraries

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/detectree2.ipynb

Import necessary modules from leafmap and samgeo.detectree2 for tree crown delineation.

```python
import leafmap
from samgeo.detectree2 import (
    TreeCrownDelineator,
    list_pretrained_models,
)
```

--------------------------------

### Download SAM 3 model

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/installation.md

Downloads the SAM 3 model from Hugging Face Hub using Pixi. Requires prior authentication.

```bash
pixi run hf download facebook/sam3
```

--------------------------------

### Define Bounding Box for Image Download

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/fast_sam.ipynb

Get the user-defined region of interest (ROI) bounds from the map. If no ROI is drawn, a default bounding box is used.

```python
bbox = m.user_roi_bounds()
if bbox is None:
    bbox = [-51.2565, -22.1777, -51.2512, -22.175]
```

--------------------------------

### Define text prompt

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/text_swimming_pools.ipynb

Sets the text prompt to be used for segmenting specific objects in the image. This prompt guides the model to identify features matching the description.

```python
text_prompt = "swimming pool"
```

--------------------------------

### Download Vector File for Box Prompts

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam2_box_prompts.ipynb

Download a sample GeoJSON file containing bounding boxes from a GitHub repository. This file will be used as input for segmentation.

```python
url = "https://github.com/opengeos/datasets/releases/download/samgeo/building_bboxes.geojson"
geojson = "building_bboxes.geojson"
leafmap.download_file(url, geojson)
```

--------------------------------

### Run segment-geospatial with GPU support via Docker

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/installation.md

This command starts a Docker container for segment-geospatial with GPU acceleration enabled. It maps port 8888 for Jupyter Notebook access.

```bash
docker run -it -p 8888:8888 --gpus=all giswqs/segment-geospatial:latest
```

--------------------------------

### Initialize SAM Predictor and Perform Segmentation

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/satellite-predictor.ipynb

Load the SAM model, initialize the SamGeoPredictor, set the input image, and run the prediction using the defined clipping box.

```python
import cv2

img_arr = cv2.imread(image)

model_type = "vit_h"

sam = sam_model_registry[model_type](checkpoint=checkpoint)

predictor = SamGeoPredictor(sam)

predictor.set_image(img_arr)

masks, _, _ = predictor.predict(src_fp=image, geo_box=clip_box)
```

--------------------------------

### Download Sample Image

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/fast_sam.ipynb

Download a GeoTIFF image for the specified bounding box using tile map services. The image is saved as 'Image.tif'.

```python
image = "Image.tif"
tms_to_geotiff(output=image, bbox=bbox, zoom=19, source="Satellite", overwrite=True)
```

--------------------------------

### Add Multiple External Masks at Once

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_video_masks.ipynb

Add multiple external masks simultaneously using the `add_masks_prompt()` method. Object IDs are automatically assigned starting from 100 to prevent conflicts.

```python
# Add all masks at once (IDs will be auto-assigned starting from 100)
sam.add_masks_prompt(external_masks)
```

--------------------------------

### Download Sample Video Data

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/workshops/AIforGood_2025.ipynb

Download a sample MP4 video file for demonstration purposes. This video will be used for object segmentation.

```python
url = "https://github.com/opengeos/datasets/releases/download/videos/cars.mp4"
```

--------------------------------

### Define Bounding Boxes for Segmentation

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_box_prompts.ipynb

Define bounding boxes either by user interaction on the map or by using default coordinates. These boxes guide the segmentation process to specific features.

```python
if m.user_rois is not None:
    boxes = m.user_rois
else:
    boxes = [
        [-117.5995, 47.6518, -117.5988, 47.652],
        [-117.5987, 47.6518, -117.5979, 47.652],
    ]
```

--------------------------------

### JSON Output Structure for Detections

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/api.md

This is an example of the JSON output structure when requesting detections. It includes image dimensions, the number of detections, and a list of detected objects with their bounding boxes and IDs.

```json
{
  "image_width": 2647,
  "image_height": 1464,
  "num_detections": 12,
  "detections": [
    {"id": 1, "value": 1, "bbox": [50, 80, 200, 250], "width": 150, "height": 170},
    {"id": 2, "value": 2, "bbox": [310, 45, 480, 210], "width": 170, "height": 165}
  ]
}
```

--------------------------------

### Initialize SAMGeo with custom options

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/automatic_mask_generator.ipynb

Re-initialize the SamGeo class, this time passing the custom `sam_kwargs` dictionary to fine-tune the mask generation process.

```python
sam = SamGeo(
    model_type="vit_h",
    sam_kwargs=sam_kwargs,
)
```

--------------------------------

### Initialize SAM3 Model

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_batch_segmentation.ipynb

Initialize the SamGeo3 model. You can choose the backend ('meta' or 'transformers') and specify device and checkpoint path if needed. `load_from_HF=True` is used for Hugging Face models.

```python
sam3 = SamGeo3(backend="meta", device=None, checkpoint_path=None, load_from_HF=True)
```

--------------------------------

### Get GeoJSON Mask Polygons

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/api.md

Use this snippet to send a GeoTIFF image to the API and receive GeoJSON mask polygons for a given prompt. Ensure the image file is opened in binary read mode.

```python
import requests

url = "http://localhost:8000/segment/text"

# Get GeoJSON mask polygons
with open("image.tif", "rb") as f:
    response = requests.post(
        url,
        files={"file": ("image.tif", f, "image/tiff")},
        data={"prompt": "building", "output_format": "geojson"},
    )

geojson = response.json()
print(f"Found {len(geojson['features'])} features")
```

--------------------------------

### Download Sample Image

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/tree_mapping.ipynb

Download a satellite image for a specified bounding box. If no bounding box is drawn by the user, a default bounding box is used. The image is saved as 'Image.tif'.

```python
bbox = m.user_roi_bounds()
if bbox is None:
    bbox = [-51.2565, -22.1777, -51.2512, -22.175]

image = "Image.tif"
leafmap.map_tiles_to_geotiff(
    output=image, bbox=bbox, zoom=19, source="Satellite", overwrite=True
)
```

--------------------------------

### Download sample image using bounding box

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/box_prompts.ipynb

Define a bounding box and download a GeoTIFF image from satellite tiles. If no bounding box is drawn by the user, a default bounding box is used.

```python
bbox = m.user_roi_bounds()
if bbox is None:
    bbox = [-51.2565, -22.1777, -51.2512, -22.175]

image = "Image.tif"
tms_to_geotiff(output=image, bbox=bbox, zoom=19, source="Satellite", overwrite=True)
```

--------------------------------

### Segment Image Using GeoJSON Bounding Boxes

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam2_box_prompts.ipynb

Perform segmentation using bounding boxes provided in a GeoJSON file. The `multimask_output` parameter is set to `False` to get a single mask per box.

```python
output_masks = "building_masks.tif"
sam.predict(
    boxes=geojson,
    point_crs="EPSG:4326",
    output=output_masks,
    dtype="uint8",
    multimask_output=False,
)
```

--------------------------------

### Add External Masks Individually

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_video_masks.ipynb

Add extracted external masks to the tracker one by one using `add_mask_prompt()`. Assign new object IDs (starting from 100) to avoid conflicts with any masks generated by text prompts.

```python
# Add masks one by one using add_mask_prompt()
# Use obj_ids starting from 100 to avoid conflicts with detected objects
for i, mask in enumerate(external_masks):
    sam.add_mask_prompt(mask, obj_id=100 + i, frame_idx=0)
```

--------------------------------

### Get user region of interest or default bounding box

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/automatic_mask_generator.ipynb

Retrieve the bounding box coordinates defined by the user on the map, or use a default bounding box if no user input is provided. This defines the area for image download.

```python
if m.user_roi_bounds() is not None:
    bbox = m.user_roi_bounds()
else:
    bbox = [-122.2659, 37.8682, -122.2521, 37.8741]
```

--------------------------------

### Download Sample Image

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam2_automatic.ipynb

Download a satellite image tile for the specified bounding box and save it as a GeoTIFF file. This function allows for overwriting existing files.

```python
image = "satellite.tif"
leafmap.map_tiles_to_geotiff(
    output=image, bbox=bbox, zoom=20, source="Satellite", overwrite=True
)
```

--------------------------------

### Add Point Prompts for Segmentation

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_video_prompts.ipynb

Add point prompts to the first frame (frame_idx=0) to guide the segmentation. Positive prompts (label=1) indicate foreground, and negative prompts (label=0) indicate background. Each object is assigned a unique obj_id.

```python
sam.add_point_prompts([[300, 200]], [1], obj_id=1, frame_idx=0)
```

```python
sam.add_point_prompts([[420, 200]], [1], obj_id=2, frame_idx=0)
```

--------------------------------

### Download Sample Data

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/maxar_open_data.ipynb

Download a sample satellite image of Derna, Libya from a provided URL. This image will be used for segmentation.

```python
url = "https://github.com/opengeos/datasets/releases/download/raster/Derna_sample.tif"
leafmap.download_file(url, output="image.tif")
```

--------------------------------

### Refine Segmentation with Multiple Points

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_video_segmentation.ipynb

Use a combination of positive and negative point prompts to precisely refine segmentation masks. This example refines object 2 to segment only the shirt, excluding the pants, by providing specific point coordinates and labels.

```python
# Refine to segment only the shirt (not pants)
sam.add_point_prompts(
    points=[[335, 195], [335, 220]],  # detect windshield, not the car
    labels=[1, 0],  # positive, negative
    obj_id=2,
    frame_idx=0,
)
sam.propagate()
sam.show_frames(frame_stride=20, ncols=3)
```

--------------------------------

### Import Libraries

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_automated_segmentation.ipynb

Imports core libraries for leafmap visualization, SAM 3, and image captioning. These are essential for the subsequent steps in the notebook.

```python
import leafmap
from samgeo import SamGeo3, download_file
from samgeo.caption import ImageCaptioner
```

--------------------------------

### Add Point Prompts for Object Refinement

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_video_segmentation.ipynb

Refine segmentation by adding point prompts. Use positive points (label 1) to include areas and negative points (label 0) to exclude them. This example adds back object 2 using a positive point click on the first frame.

```python
# Add back object 2 with a positive point click
sam.add_point_prompts(
    points=[[335, 203]],  # [x, y] coordinates
    labels=[1],  # 1=positive, 0=negative
    obj_id=2,
    frame_idx=0,
)
sam.propagate()
sam.show_frame(0)
```

--------------------------------

### Import Libraries

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_object_tracking.ipynb

Import necessary libraries for video segmentation and file handling.

```python
import os
from samgeo import SamGeo3Video, download_file
```

--------------------------------

### Run Pytest for cli-anything-samgeo

Source: https://github.com/opengeos/segment-geospatial/blob/main/agent-harness/cli_anything/samgeo/README.md

Execute the test suite for the cli-anything-samgeo module. Navigate to the agent-harness directory and run pytest with verbose and capture-off flags.

```bash
cd agent-harness
python -m pytest cli_anything/samgeo/tests/ -v -s
```

--------------------------------

### Initialize SAM 3 Video Tracker

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_video_masks.ipynb

Initialize the SAM 3 video tracker. This object will be used for segmentation and tracking tasks.

```python
# Initialize SAM 3 video predictor
sam = SamGeo3Video()
```

--------------------------------

### Configure SAM2 for Advanced Mask Generation

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam2_automatic.ipynb

Initialize SamGeo2 with advanced parameters to fine-tune automatic mask generation. This includes adjusting sampling density, confidence thresholds, and post-processing options.

```python
sam2 = SamGeo2(
    model_id="sam2-hiera-large",
    apply_postprocessing=False,
    points_per_side=32,
    points_per_batch=64,
    pred_iou_thresh=0.7,
    stability_score_thresh=0.92,
    stability_score_offset=0.7,
    crop_n_layers=1,
    box_nms_thresh=0.7,
    crop_n_points_downscale_factor=2,
    min_mask_region_area=25.0,
    use_m2m=True,
)
```

--------------------------------

### Initialize Tree Crown Delineator

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/detectree2.ipynb

Initialize the TreeCrownDelineator with a pre-trained model. The model is downloaded automatically on first use. Configure confidence and NMS thresholds.

```python
delineator = TreeCrownDelineator(
    model_name="default",  # Options: 'paracou', 'sepilok', 'danum', 'default'
    confidence_threshold=0.5,  # Minimum confidence for predictions
    nms_threshold=0.3,  # Non-maximum suppression threshold
)
```

--------------------------------

### Initialize SAM3 Model

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_image_segmentation_jpg.ipynb

Initializes the SamGeo3 model, allowing selection of the backend ('meta' or 'transformers') and specifying device and checkpoint path. `load_from_HF=True` is used when using the transformers backend.

```python
sam3 = SamGeo3(
    backend="transformers", device=None, checkpoint_path=None, load_from_HF=True
)
```

--------------------------------

### Import Libraries

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam3_batch_segmentation.ipynb

Import the necessary libraries for using leafmap and the SamGeo3 model. Ensure these are imported before proceeding.

```python
import leafmap
from samgeo import SamGeo3, download_file
```

--------------------------------

### Create and Add Layers to Leafmap

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/workshops/IPPN_2024.ipynb

Initialize a leafmap and add the original image and bounding box vector layer with custom styling.

```python
m = leafmap.Map()
m.add_raster(image, layer_name="image")
style = {
    "color": "#ffff00",
    "weight": 2,
    "fillColor": "#7c4185",
    "fillOpacity": 0,
}
m.add_vector(geojson, style=style, zoom_to_layer=True, layer_name="Bounding boxes")
```

--------------------------------

### Import necessary libraries

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/automatic_mask_generator.ipynb

Import the leafmap, SamGeo, and common utility functions for geospatial operations and SAM integration.

```python
import leafmap
from samgeo import SamGeo
from samgeo.common import overlay_images, tms_to_geotiff
```

--------------------------------

### Download Sample Video Data

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/sam2_video.ipynb

Download a sample MP4 video file for testing video segmentation.

```python
url = "https://github.com/opengeos/datasets/releases/download/videos/cars.mp4"
video_path = url
predictor.set_video(video_path)
```

--------------------------------

### Initialize SAMGeo class

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/automatic_mask_generator.ipynb

Initialize the SamGeo class with a specified model type. If the model checkpoint is not found, it will be downloaded to the current working directory.

```python
sam = SamGeo(
    model_type="vit_h",
    sam_kwargs=None,
)
```

--------------------------------

### Import necessary libraries

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/automatic_mask_generator_hq.ipynb

Imports the required modules for map creation, SAM model interaction, and geospatial utilities.

```python
import leafmap
from samgeo.hq_sam import SamGeo
from samgeo.common import overlay_images, tms_to_geotiff
```

--------------------------------

### Download vector file for box prompts

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/box_prompts.ipynb

Download a sample GeoJSON file containing bounding boxes from a GitHub URL using leafmap's download utility.

```python
url = "https://opengeos.github.io/data/sam/tree_boxes.geojson"
geojson = "tree_boxes.geojson"
leafmap.download_file(url, geojson)
```

--------------------------------

### List Available Pre-trained Models

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/examples/detectree2.ipynb

Retrieve and print the names and download URLs of available pre-trained models for detectree2.

```python
models = list_pretrained_models()
for name, url in models.items():
    print(f"{name}: {url}")
```

--------------------------------

### Pixi.toml configuration for CPU

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/installation.md

TOML configuration for a Pixi project with CPU-only support. Uses pytorch-cpu and includes libopenblas.

```toml
[workspace]
channels = ["https://prefix.dev/conda-forge"]
name = "geo"
platforms = ["linux-64", "win-64"]

[dependencies]
python = "3.12.*"
pytorch-cpu = ">=2.7.1,<3"
segment-geospatial = ">=1.2.0"
sam3 = ">=0.1.0.20251211"
jupyterlab = "*"
ipykernel = "*"
libopenblas = ">=0.3.30"
```

--------------------------------

### List Basemaps

Source: https://github.com/opengeos/segment-geospatial/blob/main/agent-harness/cli_anything/samgeo/skills/SKILL.md

List available basemap sources.

```bash
cli-anything-samgeo data basemaps [--all]
```

--------------------------------

### Authenticate with Hugging Face Hub

Source: https://github.com/opengeos/segment-geospatial/blob/main/docs/installation.md

Logs into the Hugging Face Hub using Pixi. Required before downloading models like SAM 3.

```bash
pixi run hf auth login
```