dataeval.utils.datasets.SeaDrone

class dataeval.utils.datasets.SeaDrone(root, image_set='train', transforms=None, download=False, verbose=False)

A UAV dataset focused on open water object detection.

The dataset comes from the paper SeaDronesSee: A Maritime Benchmark for Detecting Humans in Open Water by L.A. Varga et. al. (2022).

This contains the compressed (.jpg) version of the object detection v2 dataset. Dataset is 9.96 GB. Images were collected at varying altitudes (5-260 m), at a variety of viewing angles (0-90°) and drone speeds. All images contain the above metadata as well as some additional information (based on the drone), while only the training and validation images contain ground truth labels. Three different drones were used to collect the images, 2 quadcopters and 1 fixed-wing.

There are 14,227 images: 8930 images in the training set, 1547 images in the validation set, and 3750 images in the test set. There are 5 classes in the dataset: 1-swimmer, 2-boat, 3-jetski, 4-life_saving_appliances, 5-buoy. The annotation files provide the metadata and the ground-truth labels in the COCO format. Bounding boxes are (x, y, w, h). The images come in a variety of sizes, with 3632 x 5456 (149 images) or 5456 x 3632 (47 images) being the largest size, 2160 x 3840 (1030 images) being the most common size, and 932 x 1230 (5 images) being the smallest size.

Parameters:
root : str or pathlib.Path

Root directory of dataset where the milco folder exists.

image_set : "train", "val", "test", or "base", default "train"

If “base”, then the full dataset is selected (train, val and test).

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.

Type:

pathlib.Path

image_set

The selected image set from the dataset.

Type:

“train”, “val”, “test”, or “base”

index2label

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

Type:

dict[int, str]

label2index

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

Type:

dict[str, int]

Return type:

dict[str, int]

metadata

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

Type:

DatasetMetadata

transforms

The transforms to be applied to the data.

Type:

Sequence[Transform]

size

The size of the dataset.

Type:

int

Note

Data License: Creative Commons CC0 1.0 Universal