Source code for torchgeo.datasets.zuericrop
# Copyright (c) Microsoft Corporation. All rights reserved.
# Licensed under the MIT License.
"""ZueriCrop dataset."""
import os
from typing import Callable, Dict, Optional, Sequence, Tuple
import matplotlib.pyplot as plt
import torch
from torch import Tensor
from .geo import NonGeoDataset
from .utils import download_url, percentile_normalization
[docs]class ZueriCrop(NonGeoDataset):
"""ZueriCrop dataset.
The `ZueriCrop <https://github.com/0zgur0/multi-stage-convSTAR-network>`__
dataset is a dataset for time-series instance segmentation of crops.
Dataset features:
* Sentinel-2 multispectral imagery
* instance masks of 48 crop categories
* nine multispectral bands
* 116k images with 10 m per pixel resolution (24x24 px)
* ~28k time-series containing 142 images each
Dataset format:
* single hdf5 dataset containing images, semantic masks, and instance masks
* data is parsed into images and instance masks, boxes, and labels
* one mask per time-series
Dataset classes:
* 48 fine-grained hierarchical crop `categories
<https://github.com/0zgur0/multi-stage-convSTAR-network/blob/fa92b5b3cb77f5171c5c3be740cd6e6395cc29b6/labels.csv>`_
If you use this dataset in your research, please cite the following paper:
* https://doi.org/10.1016/j.rse.2021.112603
.. note::
This dataset requires the following additional library to be installed:
* `h5py <https://pypi.org/project/h5py/>`_ to load the dataset
"""
urls = [
"https://polybox.ethz.ch/index.php/s/uXfdr2AcXE3QNB6/download",
"https://raw.githubusercontent.com/0zgur0/multi-stage-convSTAR-network/fa92b5b3cb77f5171c5c3be740cd6e6395cc29b6/labels.csv", # noqa: E501
]
md5s = ["1635231df67f3d25f4f1e62c98e221a4", "5118398c7a5bbc246f5f6bb35d8d529b"]
filenames = ["ZueriCrop.hdf5", "labels.csv"]
band_names = ("NIR", "B03", "B02", "B04", "B05", "B06", "B07", "B11", "B12")
rgb_bands = ["B04", "B03", "B02"]
[docs] def __init__(
self,
root: str = "data",
bands: Sequence[str] = band_names,
transforms: Optional[Callable[[Dict[str, Tensor]], Dict[str, Tensor]]] = None,
download: bool = False,
checksum: bool = False,
) -> None:
"""Initialize a new ZueriCrop dataset instance.
Args:
root: root directory where dataset can be found
bands: the subset of bands 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:
RuntimeError: if ``download=False`` and data is not found, or checksums
don't match
"""
self._validate_bands(bands)
self.band_indices = torch.tensor(
[self.band_names.index(b) for b in bands]
).long()
self.root = root
self.bands = bands
self.transforms = transforms
self.download = download
self.checksum = checksum
self.filepath = os.path.join(root, "ZueriCrop.hdf5")
self._verify()
try:
import h5py # noqa: F401
except ImportError:
raise ImportError(
"h5py is not installed and is required to use this dataset"
)
[docs] def __getitem__(self, index: int) -> Dict[str, Tensor]:
"""Return an index within the dataset.
Args:
index: index to return
Returns:
sample containing image, mask, bounding boxes, and target label
"""
image = self._load_image(index)
mask, boxes, label = self._load_target(index)
sample = {"image": image, "mask": mask, "boxes": boxes, "label": label}
if self.transforms is not None:
sample = self.transforms(sample)
return sample
[docs] def __len__(self) -> int:
"""Return the number of data points in the dataset.
Returns:
length of the dataset
"""
import h5py
with h5py.File(self.filepath, "r") as f:
length: int = f["data"].shape[0]
return length
def _load_image(self, index: int) -> Tensor:
"""Load a single image.
Args:
index: index to return
Returns:
the image
"""
import h5py
with h5py.File(self.filepath, "r") as f:
array = f["data"][index, ...]
tensor = torch.from_numpy(array)
# Convert from TxHxWxC to TxCxHxW
tensor = tensor.permute((0, 3, 1, 2))
tensor = torch.index_select(tensor, dim=1, index=self.band_indices)
return tensor
def _load_target(self, index: int) -> Tuple[Tensor, Tensor, Tensor]:
"""Load the target mask for a single image.
Args:
index: index to return
Returns:
the target mask and label for each mask
"""
import h5py
with h5py.File(self.filepath, "r") as f:
mask_array = f["gt"][index, ...]
instance_array = f["gt_instance"][index, ...]
mask_tensor = torch.from_numpy(mask_array)
instance_tensor = torch.from_numpy(instance_array)
# Convert from HxWxC to CxHxW
mask_tensor = mask_tensor.permute((2, 0, 1))
instance_tensor = instance_tensor.permute((2, 0, 1))
# Convert instance mask of N instances to N binary instance masks
instance_ids = torch.unique(instance_tensor)
# Exclude a mask for unknown/background
instance_ids = instance_ids[instance_ids != 0]
instance_ids = instance_ids[:, None, None]
masks = instance_tensor == instance_ids
# Parse labels for each instance
labels_list = []
for mask in masks:
label = mask_tensor[mask[None, :, :]]
label = torch.unique(label)[0]
labels_list.append(label)
# Get bounding boxes for each instance
boxes_list = []
for mask in masks:
pos = torch.where(mask)
xmin = torch.min(pos[1])
xmax = torch.max(pos[1])
ymin = torch.min(pos[0])
ymax = torch.max(pos[0])
boxes_list.append([xmin, ymin, xmax, ymax])
masks = masks.to(torch.uint8)
boxes = torch.tensor(boxes_list).to(torch.float)
labels = torch.tensor(labels_list).to(torch.long)
return masks, boxes, labels
def _verify(self) -> None:
"""Verify the integrity of the dataset.
Raises:
RuntimeError: if ``download=False`` but dataset is missing or checksum fails
"""
# Check if the files already exist
exists = []
for filename in self.filenames:
filepath = os.path.join(self.root, filename)
exists.append(os.path.exists(filepath))
if all(exists):
return
# Check if the user requested to download the dataset
if not self.download:
raise RuntimeError(
"Dataset not found in `root` directory and `download=False`, "
"either specify a different `root` directory or use `download=True` "
"to automatically download the dataset."
)
# Download the dataset
self._download()
def _download(self) -> None:
"""Download the dataset."""
for url, filename, md5 in zip(self.urls, self.filenames, self.md5s):
filepath = os.path.join(self.root, filename)
if not os.path.exists(filepath):
download_url(
url,
self.root,
filename=filename,
md5=md5 if self.checksum else None,
)
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:: 0.2
"""
assert isinstance(bands, Sequence), "'bands' must be a sequence"
for band in bands:
if band not in self.band_names:
raise ValueError(f"'{band}' is an invalid band name.")
[docs] def plot(
self,
sample: Dict[str, Tensor],
time_step: int = 0,
show_titles: bool = True,
suptitle: Optional[str] = None,
) -> plt.Figure:
"""Plot a sample from the dataset.
Args:
sample: a sample returned by :meth:`__getitem__`
time_step: time step at which to access image, beginning with 0
show_titles: flag indicating whether to show titles above each panel
suptitle: optional suptitle to use for figure
Returns:
a matplotlib Figure with the rendered sample
.. 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 ValueError("Dataset doesn't contain some of the RGB bands")
ncols = 2
image, mask = sample["image"][time_step, rgb_indices], sample["mask"]
image = torch.tensor(
percentile_normalization(image.numpy()) * 255, dtype=torch.uint8
)
mask = torch.argmax(mask, dim=0)
if "prediction" in sample:
ncols += 1
preds = torch.argmax(sample["prediction"], dim=0)
fig, axs = plt.subplots(ncols=ncols, figsize=(10, 10 * ncols))
axs[0].imshow(image.permute(1, 2, 0))
axs[0].axis("off")
axs[1].imshow(mask)
axs[1].axis("off")
if show_titles:
axs[0].set_title("Image")
axs[1].set_title("Mask")
if "prediction" in sample:
axs[2].imshow(preds)
axs[2].axis("off")
if show_titles:
axs[2].set_title("Prediction")
if suptitle is not None:
plt.suptitle(suptitle)
return fig
