Source code for torchgeo.datasets.eurosat
# Copyright (c) TorchGeo Contributors. All rights reserved.
# Licensed under the MIT License.
"""EuroSAT dataset."""
import os
from collections.abc import Callable, Sequence
from typing import ClassVar, cast
import matplotlib.pyplot as plt
import numpy as np
import torch
from matplotlib.figure import Figure
from torch import Tensor
from .errors import DatasetNotFoundError, RGBBandsMissingError
from .geo import NonGeoClassificationDataset
from .utils import (
Path,
download_and_extract_archive,
download_url,
extract_archive,
rasterio_loader,
)
[docs]class EuroSAT(NonGeoClassificationDataset):
"""EuroSAT dataset.
The `EuroSAT <https://github.com/phelber/EuroSAT>`__ dataset is based on Sentinel-2
satellite images covering 13 spectral bands and consists of 10 target classes with
a total of 27,000 labeled and geo-referenced images.
Dataset format:
* rasters are 13-channel GeoTiffs
* labels are values in the range [0,9]
Dataset classes:
* Annual Crop
* Forest
* Herbaceous Vegetation
* Highway
* Industrial Buildings
* Pasture
* Permanent Crop
* Residential Buildings
* River
* Sea & Lake
This dataset uses the train/val/test splits defined in the "In-domain representation
learning for remote sensing" paper:
* https://arxiv.org/abs/1911.06721
If you use this dataset in your research, please cite the following papers:
* https://ieeexplore.ieee.org/document/8736785
* https://ieeexplore.ieee.org/document/8519248
"""
url = 'https://hf.co/datasets/torchgeo/eurosat/resolve/1ce6f1bfb56db63fd91b6ecc466ea67f2509774c/'
filename = 'EuroSATallBands.zip'
md5 = '5ac12b3b2557aa56e1826e981e8e200e'
# For some reason the class directories are actually nested in this directory
base_dir = os.path.join(
'ds', 'images', 'remote_sensing', 'otherDatasets', 'sentinel_2', 'tif'
)
splits = ('train', 'val', 'test')
split_filenames: ClassVar[dict[str, str]] = {
'train': 'eurosat-train.txt',
'val': 'eurosat-val.txt',
'test': 'eurosat-test.txt',
}
split_md5s: ClassVar[dict[str, str]] = {
'train': '908f142e73d6acdf3f482c5e80d851b1',
'val': '95de90f2aa998f70a3b2416bfe0687b4',
'test': '7ae5ab94471417b6e315763121e67c5f',
}
all_band_names = (
'B01',
'B02',
'B03',
'B04',
'B05',
'B06',
'B07',
'B08',
'B09',
'B10',
'B11',
'B12',
'B8A',
)
rgb_bands = ('B04', 'B03', 'B02')
BAND_SETS: ClassVar[dict[str, tuple[str, ...]]] = {
'all': all_band_names,
'rgb': rgb_bands,
}
[docs] def __init__(
self,
root: Path = 'data',
split: str = 'train',
bands: Sequence[str] = BAND_SETS['all'],
transforms: Callable[[dict[str, Tensor]], dict[str, Tensor]] | None = None,
download: bool = False,
checksum: bool = False,
) -> None:
"""Initialize a new EuroSAT dataset instance.
Args:
root: root directory where dataset can be found
split: one of "train", "val", or "test"
bands: a sequence of band names to load
transforms: a function/transform that takes input sample and its target as
entry and returns a transformed version
download: if True, download dataset and store it in the root directory
checksum: if True, check the MD5 of the downloaded files (may be slow)
Raises:
AssertionError: if ``split`` argument is invalid
DatasetNotFoundError: If dataset is not found and *download* is False.
.. versionadded:: 0.3
The *bands* parameter.
"""
self.root = root
self.split = split
self.transforms = transforms
self.download = download
self.checksum = checksum
assert self.split in {'train', 'val', 'test'}
self._validate_bands(bands)
self.bands = bands
self.band_indices = Tensor(
[self.all_band_names.index(b) for b in bands if b in self.all_band_names]
).long()
self._verify()
valid_fns = set()
with open(os.path.join(self.root, self.split_filenames[split])) as f:
for fn in f:
valid_fns.add(fn.strip().replace('.jpg', '.tif'))
def is_in_split(x: Path) -> bool:
return os.path.basename(x) in valid_fns
super().__init__(
root=os.path.join(root, self.base_dir),
transforms=transforms,
loader=rasterio_loader,
is_valid_file=is_in_split,
)
[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
"""
image, label = self._load_image(index)
image = torch.index_select(image, dim=0, index=self.band_indices).float()
sample = {'image': image, 'label': label}
if self.transforms is not None:
sample = self.transforms(sample)
return sample
def _verify(self) -> None:
"""Verify the integrity of the dataset."""
# Check split file
filename = os.path.join(self.root, self.split_filenames[self.split])
if not os.path.isfile(filename):
if self.download:
download_url(
self.url + self.split_filenames[self.split],
self.root,
md5=self.split_md5s[self.split] if self.checksum else None,
)
else:
raise DatasetNotFoundError(self)
# Check image directory
directory = os.path.join(self.root, self.base_dir)
zipfile = os.path.join(self.root, self.filename)
if os.path.isdir(directory):
return
elif os.path.isfile(zipfile):
extract_archive(zipfile)
elif self.download:
download_and_extract_archive(
self.url + self.filename,
self.root,
md5=self.md5 if self.checksum else None,
)
else:
raise DatasetNotFoundError(self)
def _validate_bands(self, bands: Sequence[str]) -> None:
"""Validate list of bands.
Args:
bands: user-provided sequence of bands to load
Raises:
AssertionError: if ``bands`` is not a sequence
ValueError: if an invalid band name is provided
.. versionadded:: 0.3
"""
assert isinstance(bands, Sequence), "'bands' must be a sequence"
for band in bands:
if band not in self.all_band_names:
raise ValueError(f"'{band}' is an invalid band name.")
[docs] def plot(
self,
sample: dict[str, Tensor],
show_titles: bool = True,
suptitle: str | None = None,
) -> Figure:
"""Plot a sample from the dataset.
Args:
sample: a sample returned by :meth:`NonGeoClassificationDataset.__getitem__`
show_titles: flag indicating whether to show titles above each panel
suptitle: optional string to use as a suptitle
Returns:
a matplotlib Figure with the rendered sample
Raises:
RGBBandsMissingError: If *bands* does not include all RGB bands.
.. versionadded:: 0.2
"""
rgb_indices = []
for band in self.rgb_bands:
if band in self.bands:
rgb_indices.append(self.bands.index(band))
else:
raise RGBBandsMissingError()
image = np.take(sample['image'].numpy(), indices=rgb_indices, axis=0)
image = np.rollaxis(image, 0, 3)
image = np.clip(image / 3000, 0, 1)
label = cast(int, sample['label'].item())
label_class = self.classes[label]
showing_predictions = 'prediction' in sample
if showing_predictions:
prediction = cast(int, sample['prediction'].item())
prediction_class = self.classes[prediction]
fig, ax = plt.subplots(figsize=(4, 4))
ax.imshow(image)
ax.axis('off')
if show_titles:
title = f'Label: {label_class}'
if showing_predictions:
title += f'\nPrediction: {prediction_class}'
ax.set_title(title)
if suptitle is not None:
plt.suptitle(suptitle)
return fig
[docs]class EuroSATSpatial(EuroSAT):
"""Overrides the default EuroSAT dataset splits.
Splits the data into training, validation, and test sets based on longitude.
The splits are distributed as 60%, 20%, and 20% respectively.
.. versionadded:: 0.6
"""
split_filenames: ClassVar[dict[str, str]] = {
'train': 'eurosat-spatial-train.txt',
'val': 'eurosat-spatial-val.txt',
'test': 'eurosat-spatial-test.txt',
}
split_md5s: ClassVar[dict[str, str]] = {
'train': '7be3254be39f23ce4d4d144290c93292',
'val': 'acf392290050bb3df790dc8fc0ebf193',
'test': '5ec1733f9c16116bf0aa2d921fc613ef',
}
[docs]class EuroSAT100(EuroSAT):
"""Subset of EuroSAT containing only 100 images.
Intended for tutorials and demonstrations, not for benchmarking.
Maintains the same file structure, classes, and train-val-test split. Each class has
10 images (6 train, 2 val, 2 test), for a total of 100 images.
.. versionadded:: 0.5
"""
filename = 'EuroSAT100.zip'
md5 = 'c21c649ba747e86eda813407ef17d596'
split_filenames: ClassVar[dict[str, str]] = {
'train': 'eurosat-100-train.txt',
'val': 'eurosat-100-val.txt',
'test': 'eurosat-100-test.txt',
}
split_md5s: ClassVar[dict[str, str]] = {
'train': '033d0c23e3a75e3fa79618b0e35fe1c7',
'val': '3e3f8b3c344182b8d126c4cc88f3f215',
'test': 'f908f151b950f270ad18e61153579794',
}
