Shortcuts

Source code for torchgeo.samplers.batch

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

"""TorchGeo batch samplers."""

import abc
from typing import Iterator, List, Optional, Tuple, Union

import torch
from rtree.index import Index, Property
from torch.utils.data import Sampler

from ..datasets import BoundingBox, GeoDataset
from .constants import Units
from .utils import _to_tuple, get_random_bounding_box, tile_to_chips


[docs]class BatchGeoSampler(Sampler[List[BoundingBox]], abc.ABC): """Abstract base class for sampling from :class:`~torchgeo.datasets.GeoDataset`. Unlike PyTorch's :class:`~torch.utils.data.BatchSampler`, :class:`BatchGeoSampler` returns enough geospatial information to uniquely index any :class:`~torchgeo.datasets.GeoDataset`. This includes things like latitude, longitude, height, width, projection, coordinate system, and time. """
[docs] def __init__(self, dataset: GeoDataset, roi: Optional[BoundingBox] = None) -> None: """Initialize a new Sampler instance. Args: dataset: dataset to index from roi: region of interest to sample from (minx, maxx, miny, maxy, mint, maxt) (defaults to the bounds of ``dataset.index``) """ if roi is None: self.index = dataset.index roi = BoundingBox(*self.index.bounds) else: self.index = Index(interleaved=False, properties=Property(dimension=3)) hits = dataset.index.intersection(tuple(roi), objects=True) for hit in hits: bbox = BoundingBox(*hit.bounds) & roi self.index.insert(hit.id, tuple(bbox), hit.object) self.res = dataset.res self.roi = roi
[docs] @abc.abstractmethod def __iter__(self) -> Iterator[List[BoundingBox]]: """Return a batch of indices of a dataset. Returns: batch of (minx, maxx, miny, maxy, mint, maxt) coordinates to index a dataset """
[docs]class RandomBatchGeoSampler(BatchGeoSampler): """Samples batches of elements from a region of interest randomly. This is particularly useful during training when you want to maximize the size of the dataset and return as many random :term:`chips <chip>` as possible. Note that randomly sampled chips may overlap. """
[docs] def __init__( self, dataset: GeoDataset, size: Union[Tuple[float, float], float], batch_size: int, length: Optional[int] = None, roi: Optional[BoundingBox] = None, units: Units = Units.PIXELS, ) -> None: """Initialize a new Sampler instance. The ``size`` argument can either be: * a single ``float`` - in which case the same value is used for the height and width dimension * a ``tuple`` of two floats - in which case, the first *float* is used for the height dimension, and the second *float* for the width dimension .. versionchanged:: 0.3 Added ``units`` parameter, changed default to pixel units .. versionchanged:: 0.4 ``length`` parameter is now optional, a reasonable default will be used Args: dataset: dataset to index from size: dimensions of each :term:`patch` batch_size: number of samples per batch length: number of samples per epoch (defaults to approximately the maximal number of non-overlapping :term:`chips <chip>` of size ``size`` that could be sampled from the dataset) roi: region of interest to sample from (minx, maxx, miny, maxy, mint, maxt) (defaults to the bounds of ``dataset.index``) units: defines if ``size`` is in pixel or CRS units """ super().__init__(dataset, roi) self.size = _to_tuple(size) if units == Units.PIXELS: self.size = (self.size[0] * self.res, self.size[1] * self.res) self.batch_size = batch_size self.length = 0 self.hits = [] areas = [] for hit in self.index.intersection(tuple(self.roi), objects=True): bounds = BoundingBox(*hit.bounds) if ( bounds.maxx - bounds.minx >= self.size[1] and bounds.maxy - bounds.miny >= self.size[0] ): if bounds.area > 0: rows, cols = tile_to_chips(bounds, self.size) self.length += rows * cols else: self.length += 1 self.hits.append(hit) areas.append(bounds.area) if length is not None: self.length = length # torch.multinomial requires float probabilities > 0 self.areas = torch.tensor(areas, dtype=torch.float) if torch.sum(self.areas) == 0: self.areas += 1
[docs] def __iter__(self) -> Iterator[List[BoundingBox]]: """Return the indices of a dataset. Returns: batch of (minx, maxx, miny, maxy, mint, maxt) coordinates to index a dataset """ for _ in range(len(self)): # Choose a random tile, weighted by area idx = torch.multinomial(self.areas, 1) hit = self.hits[idx] bounds = BoundingBox(*hit.bounds) # Choose random indices within that tile batch = [] for _ in range(self.batch_size): bounding_box = get_random_bounding_box(bounds, self.size, self.res) batch.append(bounding_box) yield batch
[docs] def __len__(self) -> int: """Return the number of batches in a single epoch. Returns: number of batches in an epoch """ return self.length // self.batch_size

© Copyright 2021, Microsoft Corporation. Revision ce4e7ef2.

Built with Sphinx using a theme provided by Read the Docs.
Read the Docs v: latest
Versions
latest
stable
v0.4.0
v0.3.1
v0.3.0
v0.2.1
v0.2.0
v0.1.1
v0.1.0
Downloads
On Read the Docs
Project Home
Builds

Free document hosting provided by Read the Docs.

Docs

Access comprehensive developer documentation for PyTorch

View Docs

Tutorials

Get in-depth tutorials for beginners and advanced developers

View Tutorials

Resources

Find development resources and get your questions answered

View Resources