estinel

Systematic Sentinel-2 imagery processing and collaborative image classification for Antarctic research

The goal of estinel is to process Sentinel-2 imagery in a systematic way, targeting specific locations of interest. This is a project of the Integrated Digital East Antarctica program at the Australian Antarctic Division.

Overview

estinel creates a catalog of Sentinel-2 imagery with:

• Original red, green, blue bands
• Multiple “view types” with different color balance stretches
• Complete workflow traceability using the {targets} package
• Interactive browser for visualization and classification
• Collaborative rating system for image quality assessment

Data Pipeline: - Sentinel-2 imagery queried from element84.com - Processed with GDAL and the R {targets} package - Hosted on the Pawsey Supercomputing Research Centre

Interactive Browser

Try it now: Catalog Browser

The browser provides:

• Multiple view modes - Single image, slider comparison, and side-by-side viewing
• Rating system - Classify images as Good/OK/Bad with keyboard shortcuts (1/2/3)
• Collaborative workflow - Export/import ratings as JSON for team review
• Smart filtering - Multi-select filters to view specific rating categories
• Playback mode - Automated slideshow with adjustable speed (0.1s to 3s)
• Location sorting - View locations by most recent imagery or alphabetically
• Rate and Next - Auto-advance to next unrated image for rapid classification

Quick Start

1. Open the browser
2. Press ? to see keyboard shortcuts
3. Use 1/2/3 to rate images (Good/OK/Bad)
4. Enable “Rate and Next” for rapid classification
5. Export your ratings and contribute them as issues to this repository

Contributing Ratings

After rating images: 1. Click the ⭐ star button to open the rating panel 2. Click “Export Ratings” to download your classifications as JSON 3. Share your ratings by opening an issue on this repository

Community ratings can be aggregated and shared via URL import for collaborative classification efforts.

Documentation

• Design Rationale - Comprehensive documentation of design decisions, alternatives considered, and rationale for the browser and classification system
• Browser Source - Single-file HTML application (built with Claude AI)

Catalog Schema

The image catalog uses a JSON format with a hierarchical structure:

{
  "locations": [
    {
      "id": "loc_001",
      "name": "Site Name - Description",
      "images": [
        {
          "id": "unique_image_id",
          "url": "path/to/image.png",
          "date": "YYYY-MM-DD",
          "thumbnail": "path/to/thumbnail.png",
          "download": "path/to/original.tif",
          "stac-item": "path/to/stac-metadata.json"
        }
      ]
    }
  ]
}

Field descriptions: - locations[] - Array of geographic sites being monitored - id - Unique identifier for location or image - name - Human-readable location name - url - Can be a string (single view) or object with multiple view types (e.g., {"true_color": "...", "false_color": "..."}) - date - ISO 8601 date string for the image acquisition - thumbnail - Low-resolution preview image - download - Link to original GeoTIFF data - stac-item - SpatioTemporal Asset Catalog metadata (optional)

Project Status

This is an active research project. The R package structure provides organization, but the primary workflow currently uses {targets} for pipeline orchestration. Development continues for a better underlying map index for the catalog.

Installation

# Not yet on CRAN
# Install development version from GitHub:
# remotes::install_github("mdsumner/estinel")

📦 Accessing the Catalog

The satellite imagery catalog is distributed via GitHub Releases.

Quick Start

# Download compressed catalog
piggyback::pb_download(
  file = "image-catalog.json.gz",
  repo = "mdsumner/estinel",
  tag = "catalog-data",
  dest = "."
)

# Read compressed JSON directly
catalog <- jsonlite::fromJSON(gzfile("image-catalog.json.gz"))

# Explore (catalog$locations is a data frame!)
nrow(catalog$locations)           # Number of locations (98)
head(catalog$locations$name)      # Location names
head(catalog$locations$purpose)   # Purpose tags (list column)
catalog$locations$name[1]         # First location name
length(catalog$locations$images[[1]])  # Images at first location

Filter by Purpose

# Find all emperor penguin colonies
emperor_sites <- catalog$locations[
  sapply(catalog$locations$purpose, function(p) "emperor" %in% p),
]
nrow(emperor_sites)  # How many emperor colonies?

# Find all research bases
bases <- catalog$locations[
  sapply(catalog$locations$purpose, function(p) "base" %in% p),
]

# Multiple purposes
heard_sites <- catalog$locations[
  sapply(catalog$locations$purpose, function(p) "heard" %in% p),
]

Alternative: Direct Download

# Download without piggyback
url <- "https://github.com/mdsumner/estinel/releases/download/catalog-data/image-catalog.json.gz"
temp_gz <- tempfile(fileext = ".json.gz")
download.file(url, temp_gz, mode = "wb")

# Read
catalog <- jsonlite::fromJSON(gzfile(temp_gz))

# Cleanup
unlink(temp_gz)

Catalog Structure

# Data frame with columns:
str(catalog$locations, max.level = 1)

# $ id      : chr [1:98] "site_12345" ...
# $ name    : chr [1:98] "Hobart" "Davis_Station" ...
# $ purpose : List of 98 (each element is character vector)
# $ images  : List of 98 (each element is data frame of images)

# Access images for a specific location
davis_images <- catalog$locations$images[[2]]  # Davis_Station
nrow(davis_images)           # Number of images
head(davis_images$date)      # Image dates
head(davis_images$url)       # View URLs (data frame)

Updated: Every 2 hours by automated pipeline
Size: ~2-5 MB compressed, ~40 MB uncompressed
Locations: 98 Antarctic sites ## Code of Conduct

Please note that the estinel project is released with a Contributor Code of Conduct. By contributing to this project, you agree to abide by its terms.

---

This tool was built collaboratively with Claude AI as part of the estinel project.