dataeval.utils.datasets.VOCDetection¶

class dataeval.utils.datasets.VOCDetection(root, image_set='train', year='2012', transforms=None, download=False, verbose=False)¶

Pascal VOC Detection Dataset.

Parameters:¶

root : str or pathlib.Path¶: Because of the structure of the PASCAL VOC datasets, the root needs to be one of 4 folders. 1) Directory containing the year of the already downloaded dataset (i.e. …/VOCdevkit/VOC2012 <-) 2) Directory to the VOCdevkit folder of the already downloaded dataset (i.e. …/VOCdevkit <- /VOC2012) 3) Directory to the folder one level up from the VOCdevkit folder, data may or may not be already downloaded (i.e. … <- /VOCdevkit/VOC2012) 4) Directory to where you would like the dataset to be downloaded
image_set : "train", "val", "test", or "base", default "train"¶: If “test”, then dataset year must be “2007” or “2012”. Note that the 2012 test set does not contain annotations. If “base”, then the combined dataset of “train” and “val” is returned.
year : "2007", "2008", "2009", "2010", "2011" or "2012", default "2012"¶: The dataset year.
transforms : Transform, Sequence[Transform] or None, default None¶: Transform(s) to apply to the data.
download : bool, default False¶: If True, downloads the dataset from the internet and puts it in root directory. Class checks to see if data is already downloaded to ensure it does not create a duplicate download.
verbose : bool, default False¶: If True, outputs print statements.

path¶

Location of the folder containing the data.

year¶

The selected dataset year.

image_set¶

The selected image set from the dataset.

index2label¶

Dictionary which translates from class integers to the associated class strings.

label2index¶

Dictionary which translates from class strings to the associated class integers.

metadata¶

Typed dictionary containing dataset metadata, such as id which returns the dataset class name.

transforms¶

The transforms to be applied to the data.

size¶

The size of the dataset.

Note