Datasets Module

Datasets

Cityscapes 🌆

class datasets.cityscapes.Cityscapes(self, root, split='train', labels_type="label", transform=None, target_transform=None, transforms=None)

[source]

Parameters

• root (str) - Root directory of dataset where directory leftImg8bit and gtFine or gtCoarse are located.
• split (str) - The image split to use, train, test or val if mode="fine" otherwise train, train_extra or val.
• labels_type (str) - (unused)
• transform - A function/transform that takes in a PIL image and returns a transformed version. E.g, transforms.RandomCrop
• target_transform - A function/transform that takes in the target and transforms it.
• transforms - A function/transform that takes input sample and its target as entry and returns a transformed version.

Returns

• Cityscapes

Return type

• datasets.cityscapes.Cityscapes

Methods

__getitem__(self, index)

MiniCity 🌆

class datasets.minicity.MiniCity(self, root, split='train', labels_type="label", transform=None, target_transform=None, transforms=None)

[source]

Parameters

• root (str) - Root directory of dataset where directory leftImg8bit and gtFine are located.
• split (str) - The image split to use, train, test or val if mode=”fine” otherwise train, train_extra or val.
• labels_type (str) - (unused)
• transform - A function/transform that takes in a PIL image and returns a transformed version. E.g, transforms.RandomCrop
• target_transform - A function/transform that takes in the target and transforms it.
• transforms - A function/transform that takes input sample and its target as entry and returns a transformed version.

Returns

• MiniCity

Return type

• datasets.minicity.MiniCity

Methods

__getitem__(self, index)

Tunnel 🚘

class datasets.tunnel.Tunnel(self, root, split='train', labels_type="csv", transform=None, target_transform=None, transforms=None)

[source]

Parameters

• root (str) - Root directory of dataset where directory images and jsons or csvs are located.
• split (str) - The image split to use, train, test or val (images/train, images/test and images/val).
• labels_type (str) - Format used labels csv, json or label.
• transform - A function/transform that takes in a PIL image and returns a transformed version. E.g, transforms.RandomCrop
• target_transform - A function/transform that takes in the target and transforms it.
• transforms - A function/transform that takes input sample and its target as entry and returns a transformed version.

Returns

• Tunnel

Return type

• datasets.tunnel.Tunnel

Methods

__getitem__(self, index)

Parameters

• index - Index

Returns

• (image, target, filepath)

Return type

• tuple

__len__(self)

Returns

• Number of element in the split

Return type

• int

__read_json(self)

Private method used to read json file label when labels_type is json.

__read_csv(self)

Private method used to read json file label when labels_type is csv.

Tunnel directories architecture

-── dataset
    ├── images (all images are inside)
    │   ├ train
    │   └ test
    ├── labels (create in the fonction if it not exists)
    ├── jsons (dir with all json file )
    └── csvs (dir with all csv file )