I/O Plugin#

A plugin that contains a collection of helpful import/export utilities.

Installation#

fiftyone plugins download \
    https://github.com/voxel51/fiftyone-plugins \
    --plugin-names @voxel51/io

Refer to the main README for more information about managing downloaded plugins and developing plugins locally.

Usage#

  1. Launch the App:

import fiftyone as fo
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")
session = fo.launch_app(dataset)

  1. Press ` or click the Browse operations action to open the Operators list

  2. Select any of the operators listed below!

Operators#

import_samples#

Use this operator to import samples and/or labels to an existing dataset via any of the following methods:

  • A directory of media for which you want to create new samples

  • A glob pattern of media paths for which you want to create new samples

  • A directory containing media and/or labels in any supported format for which you want to create new samples

  • One or more file(s) containing labels to add to the existing samples in your dataset

This operator is essentially a wrapper around the following import recipes:

# Add a directory of media
dataset.add_images_dir(input_dir)
dataset.add_videos_dir(input_dir)

# Add a glob pattern of media
dataset.add_images_patt(glob_patt)
dataset.add_videos_patt(glob_patt)

# Add a directory of media and/or labels in a supported format
dataset.add_dir(
    dataset_dir="/path/to/data",
    dataset_type=fo.types.XXXX,
    label_field=YYYYY,
)

# Add labels to existing samples in a supported format
data_path = {os.path.basename(p): p for p in dataset.values("filepath")}
dataset.merge_dir(
    data_path=data_path,
    labels_path="/path/to/labels.json",
    dataset_types=fo.types.XXXX,
    label_field=YYYYY,
)

merge_samples#

You can use this operator to merge a dataset or view into another dataset.

This operator is essentially a wrapper around the merge_samples() method:

dst_dataset.merge_samples(src_samples, ...)

where the operator’s form allows you to configure the source collection, the destination dataset, and any applicable optional arguments for merge_samples().

merge_labels#

You can use this operator to merge labels from one field of a collection into another field.

This operator is essentially a wrapper around the merge_labels() method:

dataset_or_view.merge_labels(in_field, out_field)

export_samples#

You can use this operator to export your current dataset or view to disk in any supported format.

This operator is essentially a wrapper around the following export recipe:

# The directory to which to write the export
export_dir = "/path/for/export"

# The type of dataset to export
dataset_type = fo.types.COCODetectionDataset  # for example

# The name of the sample field containing the labels to export
label_field = "ground_truth"

# Whether to include the media files
export_media = True

dataset_or_view.export(
    export_dir=export_dir,
    dataset_type=dataset_type,
    label_field=label_field,
    export_media=export_media,
)

where the operator’s form allows you to configure the export location, dataset type, and necessary label field(s), if applicable.

draw_labels#

You can use this operator to render annotated versions of the media in a collection with the specified label data overlaid to a directory on disk.

This operator is essentially a wrapper around the draw_labels() method:

dataset_or_view.draw_labels(
    output_dir,
    label_fields=label_fields,
    ...
)

where the operator’s form allows you to configure the output directory on disk, the label field(s) to render, and any other optional arguments for draw_labels().