# Copyright (c) 2025 GeoBenchV2. All rights reserved.
# Licensed under the Apache License 2.0.
"""SpaceNet7 dataset."""
from collections.abc import Sequence
from pathlib import Path
from typing import Literal, cast
import numpy as np
import rasterio
import torch
import torch.nn as nn
from shapely import wkt
from torch import Tensor
from .base import GeoBenchBaseDataset
from .normalization import ZScoreNormalizer
from .sensor_util import DatasetBandRegistry
[docs]
class GeoBenchSpaceNet7(GeoBenchBaseDataset):
"""GeoBench version of SpaceNet7 dataset.
Multi-temporal building segmentation and tracking dataset using PlanetScope imagery,
with fine grained building footprint masks.
If you use this dataset in your research, please cite the following paper:
* https://openaccess.thecvf.com/content/CVPR2021/html/Van_Etten_The_Multi-Temporal_Urban_Development_SpaceNet_Dataset_CVPR_2021_paper.html
"""
url = "https://hf.co/datasets/aialliance/spacenet7/resolve/main/{}"
paths = ["geobench_spacenet7.tortilla"]
sha256str = ["f202abe270b729f7f2651de64cb5c6b41c5f9915109ec12b6c467afa2abcb5b6"]
dataset_band_config = DatasetBandRegistry.SPACENET7
normalization_stats = {
"means": {
"red": 116.94474029541016,
"green": 103.55889129638672,
"blue": 76.77427673339844,
"nir": 0.0,
},
"stds": {
"red": 61.655845642089844,
"green": 49.64897537231445,
"blue": 45.88066864013672,
"nir": 255.0,
},
}
band_default_order = ("red", "green", "blue")
classes = ("background", "no-building", "building")
num_classes = len(classes)
valid_metadata = ("lat", "lon")
[docs]
def __init__(
self,
root: Path,
split: Literal["train", "val", "validation", "test"],
band_order: list[str] = band_default_order,
data_normalizer: type[nn.Module] = ZScoreNormalizer,
transforms: nn.Module = None,
metadata: Sequence[str] | None = None,
download: bool = False,
) -> None:
"""Initialize SpaceNet7 dataset.
Args:
root: Path to the dataset root directory
split: The dataset split, supports 'train', 'validation', 'test'
band_order: The order of bands to return, defaults to ['red', 'green', 'blue'], if one would
specify ['red', 'green', 'blue', 'blue', 'blue'], the dataset would return images with 5 channels
in that order. This is useful for models that expect a certain band order, or
test the impact of band order on model performance.
data_normalizer: The data normalizer to apply to the data, defaults to :class:`data_util.ZScoreNormalizer`,
transforms: The transforms to apply to the data, defaults to None
metadata: metadata names to be returned as part of the sample in the
__getitem__ method. If None, no metadata is returned.
download: Whether to download the dataset
"""
split_norm: Literal["train", "validation", "test"]
if split == "val":
split_norm = "validation"
else:
split_norm = cast(Literal["train", "validation", "test"], split)
super().__init__(
root=root,
split=split_norm,
band_order=band_order,
data_normalizer=data_normalizer,
transforms=transforms,
metadata=metadata,
download=download,
)
# TODO how to setup for time-series prediction
[docs]
def __getitem__(self, index: int) -> dict[str, Tensor]:
"""Return an index within the dataset.
Args:
index: index to return
Returns:
data and label at that index
"""
sample: dict[str, Tensor] = {}
sample_row = self.data_df.read(index)
image_path = sample_row.read(0)
mask_path = sample_row.read(1)
with rasterio.open(image_path) as img_src, rasterio.open(mask_path) as mask_src:
image: np.ndarray = img_src.read(out_dtype="float32")
mask: np.ndarray = mask_src.read()
image = torch.from_numpy(image).float()
# add 1 to mask to have a true background class
mask = torch.from_numpy(mask).long().squeeze(0) + 1
image = self.rearrange_bands(image, self.band_order)
image = self.data_normalizer(image)
sample.update(image)
sample["mask"] = mask
if self.transforms is not None:
sample = self.transforms(sample)
point = wkt.loads(sample_row.iloc[0]["stac:centroid"])
lon, lat = point.x, point.y
if "lon" in self.metadata:
sample["lon"] = torch.tensor(lon)
if "lat" in self.metadata:
sample["lat"] = torch.tensor(lat)
return sample