Datasets

DirectoryParsingLabelDataset

class chainercv.datasets.DirectoryParsingLabelDataset(root, check_img_file=None, color=True, numerical_sort=False)

A label dataset whose label names are the names of the subdirectories.

The label names are the names of the directories that locate a layer below the root directory. All images locating under the subdirectoies will be categorized to classes with subdirectory names. An image is parsed only when the function check_img_file returns True by taking the path to the image as an argument. If check_img_file is None, the path with any image extensions will be parsed.

Example

A directory structure should be one like below.

root
|-- class_0
|   |-- img_0.png
|   |-- img_1.png
|
--- class_1
    |-- img_0.png
>>> from chainercv.datasets import DirectoryParsingLabelDataset
>>> dataset = DirectoryParsingLabelDataset('root')
>>> dataset.paths
['root/class_0/img_0.png', 'root/class_0/img_1.png',
'root_class_1/img_0.png']
>>> dataset.labels
array([0, 0, 1])
Parameters:
  • root (str) – The root directory.
  • check_img_file (callable) – A function to determine if a file should be included in the dataset.
  • color (bool) – If True, this dataset read images as color images.
  • numerical_sort (bool) – Label names are sorted numerically. This means that label 2 is before label 10, which is not the case when string sort is used. Regardless of this option, string sort is used for the order of files with the same label. The default value is False.

directory_parsing_label_names

chainercv.datasets.directory_parsing_label_names(root, numerical_sort=False)

Get label names from the directories that are named by them.

The label names are the names of the directories that locate a layer below the root directory.

The label names can be used together with chainercv.datasets.DirectoryParsingLabelDataset. The index of a label name corresponds to the label id that is used by the dataset to refer the label.

Parameters:
  • root (str) – The root directory.
  • numerical_sort (bool) – Label names are sorted numerically. This means that label 2 is before label 10, which is not the case when string sort is used. The default value is False.
Retruns:
list of strings: Sorted names of classes.

TransformDataset

TransformDataset

class chainercv.datasets.TransformDataset(dataset, transform)

Dataset that indexes data of a base dataset and transforms it.

This dataset wraps a base dataset by modifying the behavior of the base dataset’s __getitem__(). Arrays returned by __getitem__() of the base dataset with an integer index are transformed by the given function transform.

The function transform takes, as an argument, in_data, which is output of the base dataset’s __getitem__(), and returns transformed arrays as output. Please see the following example.

>>> from chainer.datasets import get_mnist
>>> from chainercv.datasets import TransformDataset
>>> dataset, _ = get_mnist()
>>> def transform(in_data):
>>>     img, label = in_data
>>>     img -= 0.5  # scale to [-0.5, -0.5]
>>>     return img, label
>>> dataset = TransformDataset(dataset, transform)

Note

The index used to access data is either an integer or a slice. If it is a slice, the base dataset is assumed to return a list of outputs each corresponding to the output of the integer indexing.

Note

This class is deprecated. Please use chainer.datasets.TransformDataset instead.

Parameters:
  • dataset – Underlying dataset. The index of this dataset corresponds to the index of the base dataset.
  • transform (callable) – A function that is called to transform values returned by the underlying dataset’s __getitem__().

ADE20K

ADE20KSemanticSegmentationDataset

class chainercv.datasets.ADE20KSemanticSegmentationDataset(data_dir='auto', split='train')

Semantic segmentation dataset for ADE20K.

This is ADE20K dataset distributed in MIT Scene Parsing Benchmark website. It has 20,210 training images and 2,000 validation images.

Parameters:
  • data_dir (string) – Path to the dataset directory. The directory should contain the ADEChallengeData2016 directory. And that directory should contain at least images and annotations directries. If auto is given, the dataset is automatically downloaded into $CHAINER_DATASET_ROOT/pfnet/chainercv/ade20k.
  • split ({'train', 'val'}) – Select from dataset splits used in MIT Scene Parsing Benchmark dataset (ADE20K).

ADE20KTestImageDataset

class chainercv.datasets.ADE20KTestImageDataset(data_dir='auto')

Image dataset for test split of ADE20K.

This is an image dataset of test split in ADE20K dataset distributed at MIT Scene Parsing Benchmark website. It has 3,352 test images.

Parameters:data_dir (string) – Path to the dataset directory. The directory should contain the release_test dir. If auto is given, the dataset is automatically downloaded into $CHAINER_DATASET_ROOT/pfnet/chainercv/ade20k.

CamVid

CamVidDataset

class chainercv.datasets.CamVidDataset(data_dir='auto', split='train')

Semantic segmentation dataset for CamVid u.

Parameters:
  • data_dir (string) – Path to the root of the training data. If this is auto, this class will automatically download data for you under $CHAINER_DATASET_ROOT/pfnet/chainercv/camvid.
  • split ({'train', 'val', 'test'}) – Select from dataset splits used in CamVid Dataset.

CityscapesSemanticSegmentationDataset

class chainercv.datasets.CityscapesSemanticSegmentationDataset(data_dir=None, label_resolution=None, split='train', ignore_labels=True)

Semantic segmentation dataset for Cityscapes dataset.

Note

Please manually downalod the data because it is not allowed to re-distribute Cityscapes dataset.

Parameters:
  • data_dir (string) – Path to the dataset directory. The directory should contain at least two directories, leftImg8bit and either gtFine or gtCoarse. If None is given, it uses $CHAINER_DATSET_ROOT/pfnet/chainercv/cityscapes by default.
  • label_resolution ({'fine', 'coarse'}) – The resolution of the labels. It should be either fine or coarse.
  • split ({'train', 'val'}) – Select from dataset splits used in Cityscapes dataset.
  • ignore_labels (bool) – If True, the labels marked ignoreInEval defined in the original cityscapesScripts<https://github.com/mcordts/cityscapesScripts>_ will be replaced with -1 in the get_example() method. The default value is True.

CUB

CUBLabelDataset

class chainercv.datasets.CUBLabelDataset(data_dir='auto', return_bb=False, prob_map_dir='auto', return_prob_map=False)

Caltech-UCSD Birds-200-2011 dataset with annotated class labels.

When queried by an index, this dataset returns a corresponding img, label, a tuple of an image and class id. The image is in RGB and CHW format. The class id are between 0 and 199. If return_bb = True, a bounding box bb is appended to the tuple. If return_prob_map = True, a probability map prob_map is appended.

A bounding box is a one-dimensional array of shape \((4,)\). The elements of the bounding box corresponds to (y_min, x_min, y_max, x_max), where the four attributes are coordinates of the top left and the bottom right vertices. This information can optionally be retrieved from the dataset by setting return_bb = True.

The probability map of a bird shows how likely the bird is located at each pixel. If the value is close to 1, it is likely that the bird locates at that pixel. The shape of this array is \((H, W)\), where \(H\) and \(W\) are height and width of the image respectively. This information can optionally be retrieved from the dataset by setting return_prob_map = True.

Parameters:
  • data_dir (string) – Path to the root of the training data. If this is auto, this class will automatically download data for you under $CHAINER_DATASET_ROOT/pfnet/chainercv/cub.
  • return_bb (bool) – If True, this returns a bounding box around a bird. The default value is False.
  • prob_map_dir (string) – Path to the root of the probability maps. If this is auto, this class will automatically download data for you under $CHAINER_DATASET_ROOT/pfnet/chainercv/cub.
  • return_prob_map (bool) – Decide whether to include a probability map of the bird in a tuple served for a query. The default value is False.

CUBKeypointDataset

class chainercv.datasets.CUBKeypointDataset(data_dir='auto', return_bb=False, prob_map_dir='auto', return_prob_map=False)

Caltech-UCSD Birds-200-2011 dataset with annotated keypoints.

An index corresponds to each image.

When queried by an index, this dataset returns the corresponding img, keypoint, kp_mask, a tuple of an image, keypoints and a keypoint mask that indicates visible keypoints in the image. The data type of the three elements are float32, float32, bool. If return_bb = True, a bounding box bb is appended to the tuple. If return_prob_map = True, a probability map prob_map is appended.

keypoints are packed into a two dimensional array of shape \((K, 2)\), where \(K\) is the number of keypoints. Note that \(K=15\) in CUB dataset. Also note that not all fifteen keypoints are visible in an image. When a keypoint is not visible, the values stored for that keypoint are undefined. The second axis corresponds to the \(y\) and \(x\) coordinates of the keypoints in the image.

A keypoint mask array indicates whether a keypoint is visible in the image or not. This is a boolean array of shape \((K,)\).

A bounding box is a one-dimensional array of shape \((4,)\). The elements of the bounding box corresponds to (y_min, x_min, y_max, x_max), where the four attributes are coordinates of the top left and the bottom right vertices. This information can optionally be retrieved from the dataset by setting return_bb = True.

The probability map of a bird shows how likely the bird is located at each pixel. If the value is close to 1, it is likely that the bird locates at that pixel. The shape of this array is \((H, W)\), where \(H\) and \(W\) are height and width of the image respectively. This information can optionally be retrieved from the dataset by setting return_prob_map = True.

Parameters:
  • data_dir (string) – Path to the root of the training data. If this is auto, this class will automatically download data for you under $CHAINER_DATASET_ROOT/pfnet/chainercv/cub.
  • return_bb (bool) – If True, this returns a bounding box around a bird. The default value is False.
  • prob_map_dir (string) – Path to the root of the probability maps. If this is auto, this class will automatically download data for you under $CHAINER_DATASET_ROOT/pfnet/chainercv/cub.
  • return_prob_map (bool) – Decide whether to include a probability map of the bird in a tuple served for a query. The default value is False.

OnlineProducts

OnlineProductsDataset

class chainercv.datasets.OnlineProductsDataset(data_dir='auto', split='train')

Dataset class for Stanford Online Products Dataset.

When queried by an index, this dataset returns a corresponding img, class_id, super_class_id, a tuple of an image, a class id and a coarse level class id. Images are in RGB and CHW format. Class ids start from 0. The name of the \(l\) th coarse level class is \(l\) th element of chainercv.datasets.online_products_super_label_names.

The split selects train and test split of the dataset as done in [1]. The train split contains the first 11318 classes and the test split contains the remaining 11316 classes.

[1]Hyun Oh Song, Yu Xiang, Stefanie Jegelka, Silvio Savarese. Deep Metric Learning via Lifted Structured Feature Embedding. arXiv 2015.
Parameters:
  • data_dir (string) – Path to the root of the training data. If this is auto, this class will automatically download data for you under $CHAINER_DATASET_ROOT/pfnet/chainercv/online_products.
  • split ({'train', 'test'}) – Select a split of the dataset.

PASCAL VOC

VOCBboxDataset

class chainercv.datasets.VOCBboxDataset(data_dir='auto', split='train', year='2012', use_difficult=False, return_difficult=False)

Bounding box dataset for PASCAL VOC.

The index corresponds to each image.

When queried by an index, if return_difficult == False, this dataset returns a corresponding img, bbox, label, a tuple of an image, bounding boxes and labels. This is the default behaviour. If return_difficult == True, this dataset returns corresponding img, bbox, label, difficult. difficult is a boolean array that indicates whether bounding boxes are labeled as difficult or not.

The bounding boxes are packed into a two dimensional tensor of shape \((R, 4)\), where \(R\) is the number of bounding boxes in the image. The second axis represents attributes of the bounding box. They are (y_min, x_min, y_max, x_max), where the four attributes are coordinates of the top left and the bottom right vertices.

The labels are packed into a one dimensional tensor of shape \((R,)\). \(R\) is the number of bounding boxes in the image. The class name of the label \(l\) is \(l\) th element of chainercv.datasets.voc_bbox_label_names.

The array difficult is a one dimensional boolean array of shape \((R,)\). \(R\) is the number of bounding boxes in the image. If use_difficult is False, this array is a boolean array with all False.

The type of the image, the bounding boxes and the labels are as follows.

  • img.dtype == numpy.float32
  • bbox.dtype == numpy.float32
  • label.dtype == numpy.int32
  • difficult.dtype == numpy.bool
Parameters:
  • data_dir (string) – Path to the root of the training data. If this is auto, this class will automatically download data for you under $CHAINER_DATASET_ROOT/pfnet/chainercv/voc.
  • split ({'train', 'val', 'trainval', 'test'}) – Select a split of the dataset. test split is only available for 2007 dataset.
  • year ({'2007', '2012'}) – Use a dataset prepared for a challenge held in year.
  • use_difficult (bool) – If true, use images that are labeled as difficult in the original annotation.
  • return_difficult (bool) – If true, this dataset returns a boolean array that indicates whether bounding boxes are labeled as difficult or not. The default value is False.

VOCSemanticSegmentationDataset

class chainercv.datasets.VOCSemanticSegmentationDataset(data_dir='auto', split='train')

Semantic segmentation dataset for PASCAL VOC2012.

The class name of the label \(l\) is \(l\) th element of chainercv.datasets.voc_semantic_segmentation_label_names.

Parameters:
  • data_dir (string) – Path to the root of the training data. If this is auto, this class will automatically download data for you under $CHAINER_DATASET_ROOT/pfnet/chainercv/voc.
  • split ({'train', 'val', 'trainval'}) – Select a split of the dataset.