/OIDv6

Download single or multiple classes from the Open Images V6 dataset (OIDv6)

Primary LanguagePythonMIT LicenseMIT

Download single or multiple classes from the OIDv6

PyPI PyPI - Python Version PyPI - Implementation GitHub repo size Downloads PyPI - Status PyPI - License

Release history Documentation in Russian

Installation

pip install oidv6

Update

pip install --upgrade oidv6

Required packages

Packages Min version Current version
requests 2.23.0 PyPI
numpy 1.18.4 PyPI
pandas 1.0.4 PyPI
progressbar2 3.51.3 PyPI
opencv-contrib-python 4.2.0.34 PyPI
awscli 1.18.69 PyPI

Useful resources

Command line arguments

Argument                Type Description Valid Values
command str Boot command downloader
command str Language
Default value: en
en
ru
--dataset str The root directory for saving OIDv6
Default value: OIDv6
-
--type_data str Dataset
Default value: train
train
validation
test
all
--classes str Sequence of class names or text file -
--limit int Images Upload Limit
Default value: 0 (no limit)
From 0 to
--multi_classes bool Downloading classes in one directory No value
--yes bool Automatic download metadata No value
--no_labels bool No labeling No value
--hide_metadata bool Вывод метаданных No value
--no_clear_shell bool Do not clean the console before running No value

Examples


Note! Classes that are composed of several words should be surrounded by quotation marks (if they are passed directly to the command line). For example: "Organ (Musical Instrument)"


  1. Downloading classes (apple, banana, Kitchen & dining room table) from the train, validation and test sets with labels in semi-automatic mode and image limit = 4 (Language: Russian)

    CMD

    oidv6 downloader ru --dataset path_to_directory --type_data all --classes apple banana "Kitchen & dining room table" --limit 4
  2. Downloading training classes (cat, dog) from the train set with tags in automatic mode and image limit = 10 (Language: English)

    CMD

    oidv6 downloader en --dataset path_to_directory --type_data train --classes Cat dOg --limit 10 --yes
  3. Downloading validation classes (see text file) from the validation set with labels in automatic mode and image limit = 10 (Language: English)

    Text file

    person
    Organ (Musical Instrument)
    

    CMD

    oidv6 downloader --dataset path_to_directory --type_data validation --classes text_file_path --limit 10 --yes
  4. Downloading classes (axe, calculator) in one directory from the train, validation and test sets with labels in automatic mode and image limit = 12 (Language: English)

    CMD

    oidv6 downloader --dataset path_to_directory --type_data all --classes axe calculator --limit 12 --multi_classes --yes