Source code for torchgeo.datasets.usavars

# Copyright (c) Microsoft Corporation. All rights reserved.
# Licensed under the MIT License.

"""USAVars dataset."""

import glob
import os
from import Sequence
from typing import Callable, Optional

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import rasterio
import torch
from matplotlib.figure import Figure
from torch import Tensor

from .geo import NonGeoDataset
from .utils import download_url, extract_archive

[docs]class USAVars(NonGeoDataset): """USAVars dataset. The USAVars dataset is reproduction of the dataset used in the paper "`A generalizable and accessible approach to machine learning with global satellite imagery <>`_". Specifically, this dataset includes 1 sq km. crops of NAIP imagery resampled to 4m/px cenetered on ~100k points that are sampled randomly from the contiguous states in the USA. Each point contains three continuous valued labels (taken from the dataset released in the paper): tree cover percentage, elevation, and population density. Dataset format: * images are 4-channel GeoTIFFs * labels are singular float values Dataset labels: * tree cover * elevation * population density If you use this dataset in your research, please cite the following paper: * .. versionadded:: 0.3 """ data_url = "{}" # noqa: E501 dirname = "uar" md5 = "677e89fd20e5dd0fe4d29b61827c2456" label_urls = { "housing": data_url.format("housing.csv"), "income": data_url.format("income.csv"), "roads": data_url.format("roads.csv"), "nightlights": data_url.format("nightlights.csv"), "population": data_url.format("population.csv"), "elevation": data_url.format("elevation.csv"), "treecover": data_url.format("treecover.csv"), } split_metadata = { "train": { "url": data_url.format("train_split.txt"), "filename": "train_split.txt", "md5": "3f58fffbf5fe177611112550297200e7", }, "val": { "url": data_url.format("val_split.txt"), "filename": "val_split.txt", "md5": "bca7183b132b919dec0fc24fb11662a0", }, "test": { "url": data_url.format("test_split.txt"), "filename": "test_split.txt", "md5": "97bb36bc003ae0bf556a8d6e8f77141a", }, } ALL_LABELS = ["treecover", "elevation", "population"]
[docs] def __init__( self, root: str = "data", split: str = "train", labels: Sequence[str] = ALL_LABELS, transforms: Optional[Callable[[dict[str, Tensor]], dict[str, Tensor]]] = None, download: bool = False, checksum: bool = False, ) -> None: """Initialize a new USAVars dataset instance. Args: root: root directory where dataset can be found split: train/val/test split to load labels: list of labels to include 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 invalid labels are provided RuntimeError: if ``download=False`` and data is not found, or checksums don't match """ self.root = root assert split in self.split_metadata self.split = split for lab in labels: assert lab in self.ALL_LABELS self.labels = labels self.transforms = transforms = download self.checksum = checksum self._verify() self.files = self._load_files() self.label_dfs = { lab: pd.read_csv(os.path.join(self.root, lab + ".csv"), index_col="ID") for lab in self.labels }
[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 """ tif_file = self.files[index] id_ = tif_file[5:-4] sample = { "labels": Tensor( [self.label_dfs[lab].loc[id_][lab] for lab in self.labels] ), "image": self._load_image(os.path.join(self.root, "uar", tif_file)), "centroid_lat": Tensor([self.label_dfs[self.labels[0]].loc[id_]["lat"]]), "centroid_lon": Tensor([self.label_dfs[self.labels[0]].loc[id_]["lon"]]), } 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 """ return len(self.files)
def _load_files(self) -> list[str]: """Loads file names.""" with open(os.path.join(self.root, f"{self.split}_split.txt")) as f: files = return files def _load_image(self, path: str) -> Tensor: """Load a single image. Args: path: path to the image Returns: the image """ with as f: array: "np.typing.NDArray[np.int_]" = tensor = torch.from_numpy(array).float() return tensor 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 extracted files already exist pathname = os.path.join(self.root, "uar") csv_pathname = os.path.join(self.root, "*.csv") split_pathname = os.path.join(self.root, "*_split.txt") csv_split_count = (len(glob.glob(csv_pathname)), len(glob.glob(split_pathname))) if glob.glob(pathname) and csv_split_count == (7, 3): return # Check if the zip files have already been downloaded pathname = os.path.join(self.root, self.dirname + ".zip") if glob.glob(pathname) and csv_split_count == (7, 3): self._extract() return # Check if the user requested to download the dataset if not raise RuntimeError( f"Dataset not found in `root={self.root}` and `download=False`, " "either specify a different `root` directory or use `download=True` " "to automatically download the dataset." ) self._download() self._extract() def _download(self) -> None: """Download the dataset.""" for f_name in self.label_urls: download_url(self.label_urls[f_name], self.root, filename=f_name + ".csv") download_url(self.data_url, self.root, md5=self.md5 if self.checksum else None) for metadata in self.split_metadata.values(): download_url( metadata["url"], self.root, md5=metadata["md5"] if self.checksum else None, ) def _extract(self) -> None: """Extract the dataset.""" extract_archive(os.path.join(self.root, self.dirname + ".zip"))
[docs] def plot( self, sample: dict[str, Tensor], show_labels: bool = True, suptitle: Optional[str] = None, ) -> Figure: """Plot a sample from the dataset. Args: sample: a sample returned by :meth:`__getitem__` show_labels: flag indicating whether to show labels above panel suptitle: optional string to use as a suptitle Returns: a matplotlib Figure with the rendered sample """ image = sample["image"][:3].numpy() # get RGB inds image = np.moveaxis(image, 0, 2) fig, axs = plt.subplots(figsize=(10, 10)) axs.imshow(image) axs.axis("off") if show_labels: labels = [(lab, val) for lab, val in sample.items() if lab != "image"] label_string = "" for lab, val in labels: label_string += f"{lab}={round(val[0].item(), 2)} " axs.set_title(label_string) if suptitle is not None: plt.suptitle(suptitle) return fig

© Copyright 2021, Microsoft Corporation. Revision b9653beb.

Built with Sphinx using a theme provided by Read the Docs.
Read the Docs v: stable
On Read the Docs
Project Home

Free document hosting provided by Read the Docs.


Access comprehensive developer documentation for PyTorch

View Docs


Get in-depth tutorials for beginners and advanced developers

View Tutorials


Find development resources and get your questions answered

View Resources