More fixes

tests/datasets/test_geo.py

@@ -26,6 +26,7 @@
     NonGeoClassificationDataset,
     NonGeoDataset,
     RasterDataset,
+    Sample,
     Sentinel2,
     UnionDataset,
     VectorDataset,
@@ -46,7 +47,7 @@ def __init__(
         self.res = res
         self.paths = paths or []
 
-    def __getitem__(self, query: BoundingBox) -> dict[str, BoundingBox]:
+    def __getitem__(self, query: BoundingBox) -> Sample:
         hits = self.index.intersection(tuple(query), objects=True)
         hit = next(iter(hits))
         bounds = BoundingBox(*hit.bounds)
@@ -77,7 +78,7 @@ class CustomSentinelDataset(Sentinel2):
 
 
 class CustomNonGeoDataset(NonGeoDataset):
-    def __getitem__(self, index: int) -> dict[str, int]:
+    def __getitem__(self, index: int) -> Sample:
         return {'index': index}
 
     def __len__(self) -> int:

tests/datasets/test_splits.py

@@ -3,14 +3,14 @@
 
 from collections.abc import Sequence
 from math import floor, isclose
-from typing import Any
 
 import pytest
 from rasterio.crs import CRS
 
 from torchgeo.datasets import (
     BoundingBox,
     GeoDataset,
+    Sample,
     random_bbox_assignment,
     random_bbox_splitting,
     random_grid_cell_assignment,
@@ -49,7 +49,7 @@ def __init__(
         self._crs = crs
         self.res = res
 
-    def __getitem__(self, query: BoundingBox) -> dict[str, Any]:
+    def __getitem__(self, query: BoundingBox) -> Sample:
         hits = self.index.intersection(tuple(query), objects=True)
         hit = next(iter(hits))
         return {'content': hit.object}

tests/datasets/test_utils.py

@@ -15,7 +15,7 @@
 import torch
 from rasterio.crs import CRS
 
-from torchgeo.datasets import BoundingBox, DependencyNotFoundError
+from torchgeo.datasets import BoundingBox, DependencyNotFoundError, Sample
 from torchgeo.datasets.utils import (
     Executable,
     array_to_tensor,
@@ -393,13 +393,13 @@ def test_disambiguate_timestamp(
 
 class TestCollateFunctionsMatchingKeys:
     @pytest.fixture(scope='class')
-    def samples(self) -> list[dict[str, Any]]:
+    def samples(self) -> list[Sample]:
         return [
             {'image': torch.tensor([1, 2, 0]), 'crs': CRS.from_epsg(2000)},
             {'image': torch.tensor([0, 0, 3]), 'crs': CRS.from_epsg(2001)},
         ]
 
-    def test_stack_unbind_samples(self, samples: list[dict[str, Any]]) -> None:
+    def test_stack_unbind_samples(self, samples: list[Sample]) -> None:
         sample = stack_samples(samples)
         assert sample['image'].size() == torch.Size([2, 3])
         assert torch.allclose(sample['image'], torch.tensor([[1, 2, 0], [0, 0, 3]]))
@@ -410,13 +410,13 @@ def test_stack_unbind_samples(self, samples: list[dict[str, Any]]) -> None:
             assert torch.allclose(samples[i]['image'], new_samples[i]['image'])
             assert samples[i]['crs'] == new_samples[i]['crs']
 
-    def test_concat_samples(self, samples: list[dict[str, Any]]) -> None:
+    def test_concat_samples(self, samples: list[Sample]) -> None:
         sample = concat_samples(samples)
         assert sample['image'].size() == torch.Size([6])
         assert torch.allclose(sample['image'], torch.tensor([1, 2, 0, 0, 0, 3]))
         assert sample['crs'] == CRS.from_epsg(2000)
 
-    def test_merge_samples(self, samples: list[dict[str, Any]]) -> None:
+    def test_merge_samples(self, samples: list[Sample]) -> None:
         sample = merge_samples(samples)
         assert sample['image'].size() == torch.Size([3])
         assert torch.allclose(sample['image'], torch.tensor([1, 2, 3]))
@@ -425,13 +425,13 @@ def test_merge_samples(self, samples: list[dict[str, Any]]) -> None:
 
 class TestCollateFunctionsDifferingKeys:
     @pytest.fixture(scope='class')
-    def samples(self) -> list[dict[str, Any]]:
+    def samples(self) -> list[Sample]:
         return [
             {'image': torch.tensor([1, 2, 0]), 'crs1': CRS.from_epsg(2000)},
             {'mask': torch.tensor([0, 0, 3]), 'crs2': CRS.from_epsg(2001)},
         ]
 
-    def test_stack_unbind_samples(self, samples: list[dict[str, Any]]) -> None:
+    def test_stack_unbind_samples(self, samples: list[Sample]) -> None:
         sample = stack_samples(samples)
         assert sample['image'].size() == torch.Size([1, 3])
         assert sample['mask'].size() == torch.Size([1, 3])
@@ -446,7 +446,7 @@ def test_stack_unbind_samples(self, samples: list[dict[str, Any]]) -> None:
         assert torch.allclose(samples[1]['mask'], new_samples[0]['mask'])
         assert samples[1]['crs2'] == new_samples[0]['crs2']
 
-    def test_concat_samples(self, samples: list[dict[str, Any]]) -> None:
+    def test_concat_samples(self, samples: list[Sample]) -> None:
         sample = concat_samples(samples)
         assert sample['image'].size() == torch.Size([3])
         assert sample['mask'].size() == torch.Size([3])
@@ -455,7 +455,7 @@ def test_concat_samples(self, samples: list[dict[str, Any]]) -> None:
         assert sample['crs1'] == CRS.from_epsg(2000)
         assert sample['crs2'] == CRS.from_epsg(2001)
 
-    def test_merge_samples(self, samples: list[dict[str, Any]]) -> None:
+    def test_merge_samples(self, samples: list[Sample]) -> None:
         sample = merge_samples(samples)
         assert sample['image'].size() == torch.Size([3])
         assert sample['mask'].size() == torch.Size([3])

tests/samplers/test_batch.py

@@ -11,7 +11,7 @@
 from rasterio.crs import CRS
 from torch.utils.data import DataLoader
 
-from torchgeo.datasets import BoundingBox, GeoDataset, stack_samples
+from torchgeo.datasets import BoundingBox, GeoDataset, Sample, stack_samples
 from torchgeo.samplers import BatchGeoSampler, RandomBatchGeoSampler, Units
 
 
@@ -33,7 +33,7 @@ def __init__(self, crs: CRS = CRS.from_epsg(3005), res: float = 10) -> None:
         self._crs = crs
         self.res = res
 
-    def __getitem__(self, query: BoundingBox) -> dict[str, BoundingBox]:
+    def __getitem__(self, query: BoundingBox) -> Sample:
         return {'index': query}
 
 

tests/samplers/test_single.py

@@ -11,7 +11,7 @@
 from rasterio.crs import CRS
 from torch.utils.data import DataLoader
 
-from torchgeo.datasets import BoundingBox, GeoDataset, stack_samples
+from torchgeo.datasets import BoundingBox, GeoDataset, Sample, stack_samples
 from torchgeo.samplers import (
     GeoSampler,
     GridGeoSampler,
@@ -40,7 +40,7 @@ def __init__(self, crs: CRS = CRS.from_epsg(3005), res: float = 10) -> None:
         self._crs = crs
         self.res = res
 
-    def __getitem__(self, query: BoundingBox) -> dict[str, BoundingBox]:
+    def __getitem__(self, query: BoundingBox) -> Sample:
         return {'index': query}
 
 

torchgeo/datasets/enviroatlas.py

@@ -6,7 +6,7 @@
 import os
 import sys
 from collections.abc import Callable, Sequence
-from typing import ClassVar, cast
+from typing import Any, ClassVar, cast
 
 import fiona
 import matplotlib.pyplot as plt
@@ -347,8 +347,8 @@ def __getitem__(self, query: BoundingBox) -> Sample:
         """
         hits = self.index.intersection(tuple(query), objects=True)
         filepaths = cast(list[dict[str, str]], [hit.object for hit in hits])
-
-        sample: Sample = {'image': [], 'mask': [], 'crs': self.crs, 'bounds': query}
+        images: list[np.typing.NDArray[Any]] = []
+        masks: list[np.typing.NDArray[Any]] = []
 
         if len(filepaths) == 0:
             raise IndexError(
@@ -389,23 +389,27 @@ def __getitem__(self, query: BoundingBox) -> Sample:
                     'waterbodies',
                     'water',
                 ]:
-                    sample['image'].append(data)
+                    images.append(data)
                 elif layer in ['prior', 'prior_no_osm_no_buildings']:
                     if self.prior_as_input:
-                        sample['image'].append(data)
+                        images.append(data)
                     else:
-                        sample['mask'].append(data)
+                        masks.append(data)
                 elif layer in ['lc']:
                     data = self.raw_enviroatlas_to_idx_map[data]
-                    sample['mask'].append(data)
+                    masks.append(data)
         else:
             raise IndexError(f'query: {query} spans multiple tiles which is not valid')
 
-        sample['image'] = np.concatenate(sample['image'], axis=0)
-        sample['mask'] = np.concatenate(sample['mask'], axis=0)
+        image = torch.from_numpy(np.concatenate(images, axis=0))
+        mask = torch.from_numpy(np.concatenate(masks, axis=0))
 
-        sample['image'] = torch.from_numpy(sample['image'])
-        sample['mask'] = torch.from_numpy(sample['mask'])
+        sample: Sample = {
+            'image': image,
+            'mask': mask,
+            'crs': self.crs,
+            'bounds': query,
+        }
 
         if self.transforms is not None:
             sample = self.transforms(sample)

torchgeo/datasets/eurocrops.py

@@ -6,13 +6,13 @@
 import csv
 import os
 from collections.abc import Callable, Iterable
-from typing import Any
 
 import fiona
 import matplotlib.pyplot as plt
 import numpy as np
 from matplotlib.figure import Figure
 from rasterio.crs import CRS
+from torch import Tensor
 
 from .errors import DatasetNotFoundError
 from .geo import VectorDataset
@@ -247,9 +247,7 @@ def plot(
 
         fig, axs = plt.subplots(nrows=1, ncols=ncols, figsize=(4, 4))
 
-        def apply_cmap(
-            arr: 'np.typing.NDArray[Any]',
-        ) -> 'np.typing.NDArray[np.float64]':
+        def apply_cmap(arr: Tensor) -> 'np.typing.NDArray[np.float64]':
             # Color 0 as black, while applying default color map for the class indices.
             cmap = plt.get_cmap('viridis')
             im: np.typing.NDArray[np.float64] = cmap(arr / len(self.class_map))

torchgeo/datasets/fair1m.py

@@ -283,7 +283,7 @@ def __getitem__(self, index: int) -> Sample:
             label_path = label_path.replace('.tif', '.xml')
             voc = parse_pascal_voc(label_path)
             boxes, labels = self._load_target(voc['points'], voc['labels'])
-            sample: Sample = {'image': image, 'boxes': boxes, 'label': labels}
+            sample = {'image': image, 'boxes': boxes, 'label': labels}
 
         if self.transforms is not None:
             sample = self.transforms(sample)

torchgeo/datasets/gid15.py

@@ -139,7 +139,7 @@ def __getitem__(self, index: int) -> Sample:
             mask = self._load_target(files['mask'])
             sample: Sample = {'image': image, 'mask': mask}
         else:
-            sample: Sample = {'image': image}
+            sample = {'image': image}
 
         if self.transforms is not None:
             sample = self.transforms(sample)

torchgeo/datasets/skippd.py

@@ -134,7 +134,7 @@ def __len__(self) -> int:
 
         return num_datapoints
 
-    def __getitem__(self, index: int) -> dict[str, str | Tensor]:
+    def __getitem__(self, index: int) -> Sample:
         """Return an index within the dataset.
 
         Args:
@@ -143,7 +143,7 @@ def __getitem__(self, index: int) -> dict[str, str | Tensor]:
         Returns:
             data and label at that index
         """
-        sample: dict[str, str | Tensor] = {'image': self._load_image(index)}
+        sample: Sample = {'image': self._load_image(index)}
         sample.update(self._load_features(index))
 
         if self.transforms is not None:
@@ -176,7 +176,7 @@ def _load_image(self, index: int) -> Tensor:
         tensor = torch.from_numpy(arr).to(torch.float32)
         return tensor
 
-    def _load_features(self, index: int) -> dict[str, str | Tensor]:
+    def _load_features(self, index: int) -> Sample:
         """Load label.
 
         Args:
@@ -194,7 +194,7 @@ def _load_features(self, index: int) -> dict[str, str | Tensor]:
         path = os.path.join(self.root, f'times_{self.split}_{self.task}.npy')
         datestring = np.load(path, allow_pickle=True)[index].strftime(self.dateformat)
 
-        features: dict[str, str | Tensor] = {
+        features: Sample = {
             'label': torch.tensor(label, dtype=torch.float32),
             'date': datestring,
         }

torchgeo/datasets/sustainbench_crop_yield.py

@@ -98,7 +98,7 @@ def __init__(
         self._verify()
 
         self.images = []
-        self.features = []
+        self.features: list[Sample] = []
 
         for country in self.countries:
             image_file_path = os.path.join(
@@ -122,7 +122,7 @@ def __init__(
                 year = year_npz_file[idx]
                 ndvi = ndvi_npz_file[idx]
 
-                features = {
+                features: Sample = {
                     'label': torch.tensor(target).to(torch.float32),
                     'year': torch.tensor(int(year)),
                     'ndvi': torch.from_numpy(ndvi).to(dtype=torch.float32),

torchgeo/datasets/utils.py

@@ -65,8 +65,54 @@ class Sample(TypedDict, total=False):
     bounds: BoundingBox
     crs: CRS
 
-    # TODO: remove
+    # TODO: Additional dataset-specific keys that should be subclasses
+    images: Tensor
+    input: Tensor
     boxes: Tensor
+    bboxes: Tensor
+    masks: Tensor
+    labels: Tensor
+    prediction_masks: Tensor
+    prediction_boxes: Tensor
+    prediction_labels: Tensor
+    prediction_label: Tensor
+    prediction_scores: Tensor
+    audio: Tensor
+    points: Tensor
+    x: Tensor
+    y: Tensor
+    relative_time: Tensor
+    ocean: Tensor
+    array: Tensor
+    chm: Tensor
+    hsi: Tensor
+    las: Tensor
+    image1: Tensor
+    image2: Tensor
+    crs1: Tensor
+    crs2: Tensor
+    magnitude: Tensor
+    agb: Tensor
+    key: Tensor
+    patch: Tensor
+    geometry: Tensor
+    properties: Tensor
+    id: int
+    centroid_lat: Tensor
+    centroid_lon: Tensor
+    content: Tensor
+    year: Tensor
+    ndvi: Tensor
+    filename: str
+    category: str
+    field_ids: Tensor
+    tile_index: Tensor
+    transform: Tensor
+    src: Tensor
+    dst: Tensor
+    input_size: Tensor
+    output_size: Tensor
+    index: BoundingBox
 
 
 class Batch(Sample):
@@ -455,7 +501,7 @@ def stack_samples(samples: Iterable[Sample]) -> Batch:
 
     .. versionadded:: 0.2
     """
-    collated: dict[Any, Any] = _list_dict_to_dict_list(samples)
+    collated: Batch = _list_dict_to_dict_list(samples)
     for key, value in collated.items():
         if isinstance(value[0], Tensor):
             collated[key] = torch.stack(value)
@@ -475,7 +521,7 @@ def concat_samples(samples: Iterable[Sample]) -> Batch:
 
     .. versionadded:: 0.2
     """
-    collated: dict[Any, Any] = _list_dict_to_dict_list(samples)
+    collated: Batch = _list_dict_to_dict_list(samples)
     for key, value in collated.items():
         if isinstance(value[0], Tensor):
             collated[key] = torch.cat(value)
@@ -497,7 +543,7 @@ def merge_samples(samples: Iterable[Sample]) -> Batch:
 
     .. versionadded:: 0.2
     """
-    collated: dict[Any, Any] = {}
+    collated: Batch = {}
     for sample in samples:
         for key, value in sample.items():
             if key in collated and isinstance(value, Tensor):