FiftyOne terminology ¶

This cheat sheet introduces the key terminology in the world of FiftyOne.

The basics ¶

FiftyOne	The open-source framework, the core library, and the Python SDK.
FiftyOne App	The provided user interface for graphically viewing, filtering, and understanding your datasets. Can be launched in the browser or within notebooks.
FiftyOne Teams	The enterprise-grade suite built on top of FiftyOne for collaboration, permissioning, and working with cloud-backed media.

Other components ¶

Brain	Library of ML-powered capabilities for computation and visualization.
Dataset Zoo	Collection of common datasets available for download and loading into FiftyOne.
Model Zoo	Collection of pre-trained models available for download and inference.
Plugin	A module you can use to customize and extend the behavior of FiftyOne.
Operator	A plugin subcomponent that defines an operation that can be executed either directly by users in the App and/or internally invoked by other plugin components
Integration	A dataset, ML framework, annotation service, or other tool FiftyOne is directly compatible with.

Data model ¶

Dataset	Core data structure in FiftyOne, composed of Sample instances. Provides a consistent interface for loading images, videos, metadata, annotations, and predictions. The computer vision analog of a table of data.
Sample	The atomic elements of a Dataset that store all the information related to a given piece of data. Every sample has an associated media file. The computer vision analog of a row of tabular data.
DatasetView	A view into a Dataset, which can filter, sort, transform, etc. the dataset along various axes to obtain a desired subset of the samples.
ViewStage	A logical operation, such as filtering, matching, or sorting, which can be used to generate a DatasetView.
Field	Attributes of Sample instances that store customizable information about the samples. The computer vision analog of a column in a table.
Embedded Document Field	A collection of related fields organized under a single top-level Field, similar to a nested dictionary.
Label	Class hierarchy used to store semantic information about ground truth or predicted labels in a sample. Builtin Label types include Classification, Detections, Keypoints, and many others.
Tag	A field containing a list of strings representing relevant information. Tags can be assigned to datasets, samples, or labels.
Metadata	A special Sample field that can be automatically populated with media type-specific metadata about the raw media associated with the sample.
Aggregation	A class encapsulating the computation of an aggregate statistic about the contents of a dataset or view.

FiftyOne App ¶

Session	An instance of the FiftyOne App connected to a specific dataset, via which you can use to programmatically interact with the App.
Sample grid	The rectangular media grid that you can scroll through to quickly browse the samples in a dataset. Click on any media in the grid to open the sample modal.
Sample modal	The expanded modal that provides detailed information and visualization about an individual sample in a dataset.
Sidebar	Vertical component on left side of App that provides convenient options for filtering the dataset and toggling the visibility of fields in the sample grid.
View bar	Horizontal bar at the top of the App where you can create and compose view stages via point and click operations to filter your dataset and show only the content of interest.