Loads one jurisdictional leaf, evaluates a constant-velocity dead-reckoning baseline with the shipped ADE/FDE metrics, and prints the aligned context tensor shapes and interaction density.
# after: pip install -e . and downloading the dataset to /data/maris
python examples/quickstart.py --root /data/maris --track A --region dma --split testExpected output (Track A, DMA, test, OSM-consistent view):
loaded 14,888 OSM-consistent DMA Track A test samples
history (14888, 30, 2) future (14888, 30, 2)
dead-reckoning ADE = ... m FDE = ... m
env masks (64, 6, 128, 128) shore-SDF (64, 128, 128)
social: 20.5% of samples have ≥1 neighbour within 3 km
Options: --leaf <dir> points directly at a single extracted leaf;
--no-context skips the environment/social tensors; --track {A,B},
--region {dma,noaa,piraeus,norway}, --split {train,val,test} select the cut.