geosam

The geosam R package brings Meta's Segment Anything Model 3 (SAM3) to R for detecting objects in satellite imagery and photos. Describe what you're looking for in plain text—no training data or model fine-tuning required. The package is inspired by the Python package segment-geospatial by Qiusheng Wu, and aims to bring similar functionality to R users.

Install from R-Universe:

install.packages("geosam", repos = c("https://walkerke.r-universe.dev", "https://cloud.r-project.org"))

Then set up the Python environment:

library(geosam)
geosam_install()

You'll need a HuggingFace account with access to the gated SAM3 model. Run geosam_status() after installation to verify that your token can access the model metadata before downloading weights.

Read through these vignettes to learn how to use the package:

• Getting started with geosam

• Satellite imagery detection

• Regular image detection

• Interactive workflows

Support and how to learn more

If you find this project useful in your work and would like to ensure continued development of the package, you can provide support in the following ways:

• Chip in some funds to support package development via PayPal;
• Set up a consulting engagement or workshop through Walker Data to help you implement geosam in your project. Send a note to kyle@walker-data.com if you are interested;
• File an issue - or even better, a pull request - at https://github.com/walkerke/geosam/issues.

To stay on top of package updates / new features and to get information about trainings, be sure to sign up for the Walker Data mailing list here.