README.md

Figure regeneration

These scripts regenerate the data-driven figures of the MARIS-Forecast Data Descriptor directly from the released dataset. They use Arial-metric typography via Arimo (Apache-2.0, bundled in fonts/), loaded by figstyle.py; the fonts are embedded into every output PDF, so the figures are self-contained.

Setup

pip install -e ".[figures]"          # matplotlib + scipy

Each script has a DATA path constant near the top pointing at a local copy of the dataset — edit it to your download location before running. Outputs are written to figures/out/.

Scripts → figures

Script	Figures in the paper
plot_composition_geographic_transfer.py	Fig. 7 cross-region composition (vessel class / speed / displacement / scene), Fig. 8 geographic coverage, Fig. 9 cross-domain transfer heatmap
plot_environment.py	Fig. 3 paired environmental representations (raster · signed-distance field · scene descriptor)
plot_social.py	Fig. 4 social-neighbourhood context + interaction statistics

python figures/plot_composition_geographic_transfer.py
python figures/plot_environment.py
python figures/plot_social.py

Notes:

• The environment and geographic scripts flip raster/SDF grids vertically (arr[::-1]) because the release stores them north-in-row-0.
• The composition speed panel is a KDE, the displacement panel is a histogram + KDE with a share-of-samples (%) axis; the transfer heatmap uses a fixed colour scale with luminance-adaptive cell text.