Datasets:
Add EnvShip-Voyage: long-term ship trajectory imputation dataset + full build pipeline + docs
b61a1c7 verified | """Streaming environment loader for EnvShip-Voyage. | |
| Reads the shared regional basemap (Zarr) lazily and samples per-point environmental features along | |
| any trajectory — the streaming access pattern that replaces per-sample rasters. Also loads the | |
| fairway distance field / network when present. | |
| env = EnvLoader(basemap_dir) | |
| df = env.sample_track(lat_array, lon_array) # -> DataFrame of env features per point | |
| """ | |
| from __future__ import annotations | |
| import json | |
| from pathlib import Path | |
| import numpy as np | |
| class EnvLoader: | |
| def __init__(self, basemap_dir: str | Path): | |
| import zarr | |
| from pyproj import Transformer | |
| self.dir = Path(basemap_dir) | |
| self.meta = json.loads((self.dir / "basemap_meta.json").read_text()) | |
| a, b, c, d, e, f = self.meta["transform"] | |
| self.transform = (a, b, c, d, e, f) # affine: x = a*col + b*row + c ; y = d*col + e*row + f | |
| self.root = zarr.open_group(str(self.dir / "basemap.zarr"), mode="r") | |
| self.to_metric = Transformer.from_crs("EPSG:4326", self.meta["crs"], always_xy=True) | |
| self.H, self.W = self.meta["height"], self.meta["width"] | |
| self._layers = {name: self.root[name] for name in self.meta["layers"]} | |
| # fairway distance field (optional, added by S6) | |
| fd = self.dir / "fairway_dist.zarr" | |
| self.fairway = zarr.open_array(str(fd), mode="r") if fd.exists() else None | |
| def _rowcol(self, lat, lon): | |
| x, y = self.to_metric.transform(np.asarray(lon), np.asarray(lat)) | |
| a, b, c, d, e, f = self.transform | |
| # invert affine (b,d assumed 0 for north-up metric grid) | |
| col = np.round((x - c) / a).astype(int) | |
| row = np.round((y - f) / e).astype(int) | |
| inb = (col >= 0) & (col < self.W) & (row >= 0) & (row < self.H) | |
| return row, col, inb | |
| def sample_track(self, lat, lon): | |
| import pandas as pd | |
| lat = np.asarray(lat, dtype=float) | |
| lon = np.asarray(lon, dtype=float) | |
| ok = np.isfinite(lat) & np.isfinite(lon) | |
| row = np.full(lat.shape, -1) | |
| col = np.full(lat.shape, -1) | |
| if ok.any(): | |
| r, c, inb = self._rowcol(lat[ok], lon[ok]) | |
| # points outside the basemap grid are invalid (NOT clamped -> no edge artifacts) | |
| idx = np.where(ok)[0] | |
| row[idx] = r | |
| col[idx] = c | |
| ok[idx[~inb]] = False | |
| row[idx[~inb]] = -1 | |
| col[idx[~inb]] = -1 | |
| out = {} | |
| for name, arr in self._layers.items(): | |
| vals = np.full(lat.shape, np.nan, dtype="float32") | |
| if ok.any(): | |
| vals[ok] = np.asarray(arr[row[ok], col[ok]], dtype="float32") | |
| out[name] = vals | |
| if self.fairway is not None and ok.any(): | |
| fv = np.full(lat.shape, np.nan, dtype="float32") | |
| fv[ok] = np.asarray(self.fairway[row[ok], col[ok]], dtype="float32") | |
| out["dist_fairway_m"] = fv | |
| # convenience aliases | |
| res = pd.DataFrame(out) | |
| res.rename(columns={"sdf_shore": "d_shore_m", "sdf_nav": "d_nav_m", "depth": "depth_m", | |
| "landwater": "is_water"}, inplace=True) | |
| if "dist_fairway_m" in res: | |
| res["in_fairway"] = (res["dist_fairway_m"] <= 200).astype("int8") | |
| return res | |