Utilities Plugin

A plugin that contains common utilities for working with FiftyOne.

Installation

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

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)

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

3. Select any of the operators listed below!

Operators

create_dataset

You can use this operator to create a new dataset from within the App.

This operator is essentially a wrapper around the Dataset() constructor:

dataset = fo.Dataset(name, persistent=persistent)

where the operator’s form allows you to configure the name and persistence of the dataset.

load_dataset

You can use this operator to switch to a different dataset in the App.

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

session.dataset = fo.load_dataset(name)

where the operator’s form allows you to configure the name of the dataset to load.

edit_dataset_info

You can use this operator to edit dataset-level metadata in the App.

This operator is essentially a wrapper around the various dataset properties described in this section of the docs:

# Basic
dataset.name = ...
dataset.description = ...
dataset.persistent = ...
dataset.tags = ...

# Info
dataset.info = ...

# App config
dataset.app_config = ...

# Classes
dataset.classes = ...
dataset.default_classes = ...

# Mask targets
dataset.mask_targets = ...
dataset.default_mask_targets = ...

# Keypoint skeletons
dataset.skeletons = ...
dataset.default_skeleton = ...

where the operator’s form allows you to edit any of the above properties.

rename_dataset

You can use this operator to rename a dataset in the App.

This operator is essentially a wrapper around setting the dataset’s name property:

dataset.name = new_name

where the operator’s form allows you to configure the new name for the dataset.

clone_dataset

You can use this operator to clone a dataset or your current view in the App.

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

new_dataset = dataset.clone(new_name)

view = dataset.filter_labels(...)
new_dataset = view.clone(new_name)

where the operator’s form allows you to configure what dataset/view to clone.

delete_dataset

You can use this operator to delete a dataset in the App.

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

fo.delete_dataset(name)

where the operator’s form allows you to configure the name of the dataset to delete.

delete_samples

You can use this operator to delete samples from a dataset in the App.

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

# Delete the currently selected samples
dataset.delete_samples(session.selected)

# Delete the current view
datast.delete_samples(session.view)

# Delete all samples
dataset.clear()

where the operator’s form allows you to choose which samples to delete.

apply_saved_view

You can use this operator to apply a saved view from another dataset to your current dataset (assuming the view is applicable to both).

This operator is essentially a wrapper around the following code snippet:

src_view = src_dataset.load_saved_view(name)

# Apply source view to current dataset
view = fo.DatasetView._build(dataset, src_view._serialize())

where the operator’s form allows you to choose the source dataset and view to apply.

reload_saved_view

You can use this operator to reload saved views on a dataset that are generated (materialized, patches, frames, or clips).

This operator is essentially a wrapper around the following pattern:

view = dataset.load_saved_view(name)
view.reload()

dataset.save_view(name, view, overwrite=True)

where the operator’s form allows you to select the saved view to reload.

compute_metadata

You can use this operator to populate the metadata field of a collection.

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

dataset_or_view.compute_metadata(...)

where the operator’s form allows you to configure any applicable optional arguments for compute_metadata().

generate_thumbnails

You can use this operator to generate thumbnails for the media in a collection.

This operator is essentially a wrapper around the following pattern:

import fiftyone.utils.image as foui

foui.transform_images(
    dataset_or_view,
    size=(width, height),
    output_field="thumbnail_path",
    output_dir="/path/for/thumbnails",
    skip_failures=True,
)

dataset.app_config.media_fields.append("thumbnail_path")
dataset.app_config.grid_media_field = "thumbnail_path"
dataset.save()

where the operator’s form allows you to configure the (width, height) for the thumbnails, the field in which to store their paths, and the directory in which to store the thumbnail images.

delegate (SDK-only)

You can use this operator to programmatically delegate execution of an arbitrary function call that can be expressed in any of the following forms:

• Execute an arbitrary function: fcn(*args, **kwargs)

• Apply a function to a dataset or view: fcn(dataset_or_view, *args, **kwargs)

• Call an instance method of a dataset or view: dataset_or_view.fcn(*args, **kwargs)

import fiftyone as fo
import fiftyone.operators as foo
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart")
delegate = foo.get_operator("@voxel51/utils/delegate")

# Compute metadata
delegate("compute_metadata", dataset=dataset)

# Compute visualization
delegate(
    "fiftyone.brain.compute_visualization",
    dataset=dataset,
    brain_key="img_viz",
)

# Export a view
delegate(
    "export",
    view=dataset.to_patches("ground_truth"),
    export_dir="/tmp/patches",
    dataset_type="fiftyone.types.ImageClassificationDirectoryTree",
    label_field="ground_truth",
)

# Load the exported patches into a new dataset
delegate(
    "fiftyone.Dataset.from_dir",
    dataset_dir="/tmp/patches",
    dataset_type="fiftyone.types.ImageClassificationDirectoryTree",
    label_field="ground_truth",
    name="patches",
    persistent=True,
)