lunadem 0.3.0

Lunar DEM generation, Kaguya scene analysis, metadata prediction, and rover-aware landing utilities.

lunadem

lunadem is a lunar analysis toolkit for shape-from-shading DEM reconstruction, Kaguya scene metadata parsing, packaged metadata prediction, rover-aware landing safety scoring, and interactive 3D visualization.

PyPI: https://pypi.org/project/lunadem/

Install

pip install lunadem

Optional extras:

pip install "lunadem[ml]"
pip install "lunadem[viz]"
pip install "lunadem[pds]"
pip install "lunadem[dev]"

Canonical Python import is now lunadem.

Backward compatibility:

• import lunardem still works, but it is deprecated.
• lunardem remains available as a compatibility CLI alias.

What It Includes

• DEM generation from image input with sfs, multiscale_sfs, ml, and hybrid
• STAC + camera.json loaders for the bundled Kaguya dataset
• Deterministic metadata derivation helpers for geometry, timing, coverage, and coordinate conversion
• Packaged SFS explainers via CLI and Python
• Packaged metadata prediction APIs with ONNX-ready runtime hooks and baseline fallback metadata priors
• Rover preset library for Sojourner, Spirit, Opportunity, Curiosity, Perseverance, Pragyan, and SORA-Q
• Safe landing-site selection from an image or DEM
• Interactive Plotly visualizations for surfaces, landing sites, and Moon/footprint/camera/sun geometry
• download --test to fetch the official reference scene

Quick Start

from lunadem import (
    ReconstructionConfig,
    find_safe_landing_site,
    generate_dem,
    load_kaguya_scene,
    predict_scene_metadata,
    plot_scene_geometry_3d,
)

scene = load_kaguya_scene("TC1S2B0_01_07496N087E3020")

cfg = ReconstructionConfig(
    output={"output_dir": "output", "base_name": "moon_run"},
)
dem_result = generate_dem(scene.image_path, method="hybrid", config=cfg)

metadata_prediction = predict_scene_metadata(scene.image_path)
landing_site = find_safe_landing_site(scene.image_path, rover="pragyan", scene=scene)

figure = plot_scene_geometry_3d(scene, save_path="output/scene_geometry.html")

print(dem_result.exports)
print(metadata_prediction.targets)
print(landing_site.summary)

CLI Examples

lunadem generate dataset/TC1S2B0_01_07496N087E3020/image/image.tif --method hybrid --output output
lunadem scene-summary TC1S2B0_01_07496N087E3020
lunadem sfs --maths
lunadem predict dataset/TC1S2B0_01_07496N087E3020/image/image.tif --kind all
lunadem landing-site dataset/TC1S2B0_01_07496N087E3020/image/image.tif --rover pragyan --scene TC1S2B0_01_07496N087E3020
lunadem download --test --output downloads

Metadata Suite Accuracy

The bundled 18-scene Kaguya metadata suite currently ships baseline packaged statistics and ONNX-ready runtime hooks. The table below shows mean absolute error against the local dataset targets used by the packaged metrics file.

Target	MAE
sun_azimuth_deg	1.5308
sun_elevation_deg	0.0022
off_nadir_deg	0.1052
view_azimuth_deg	0.0513
gsd_m	0.0808
centroid_lat_deg	3.0003
centroid_lon_deg	0.5092

Important note:

• The current dataset is small and scene-level labels are limited.
• These predictors estimate scene metadata, not ground-truth DEM quality or landing safety labels.
• Running tools/train_metadata_cnns.py in a healthy Python environment will replace the packaged baseline with exported ONNX models and refreshed metrics.

Main Public APIs

• generate_dem(...)
• analyze_dem(...)
• assess_landing(...)
• load_stac_item(...)
• load_camera_model(...)
• load_kaguya_scene(...)
• summarize_scene_metadata(...)
• predict_illumination(...)
• predict_view_geometry(...)
• predict_scene_location(...)
• predict_scene_metadata(...)
• get_rover_spec(...)
• find_safe_landing_site(...)
• plot_3d_surface_interactive(...)
• plot_landing_site_2d(...)
• plot_landing_site_3d(...)
• plot_scene_geometry_3d(...)

Reference Files

• sfs.md
• lunar_data.md
• run.md
• update.md

License

MIT License. See LICENSE.