Quick References

Admin (organization role)

A user who can manage projects and members, similar to an owner.

Annotation

A label (e.g., bounding boxes, tags, text) that can be queried.

Annotation Manager (project role)

Manages annotation and QA tasks, redistributes assignments, and reviews annotation quality.

Annotator (project role)

Works on assigned annotation or QA tasks only, without permissions to view or manage others.

Application

Software built using the Function-as-a-Service (FaaS) model, where functionality is implemented as independent functions.

Assignment

A subset of items from the labeling task that are distributed to a specified annotator.

Attribute

Metadata about annotations (e.g., Yes/No answers, sliders, free text), providing more context.

Bot

A system-generated user used to run services. Bots log in automatically and make API calls with tokens.

Cycle

A pipeline cycle representing the full sequence of node executions during a single pipeline run, typically tied to a specific item.

Dataset

A collection of items (files) with metadata and annotations. Acts as the main unit of data storage and management.

Developer (project role)

Can manage datasets, create tasks, set recipes, and export data. Typically handles technical aspects of a project.

DQL (Dataloop Query Language)

A query language used to filter, sort, and search data items and annotations across datasets.

Execution

Functions are the basic units of an application. Defined in a class, they can be executed once the service is deployed.

Filter

A tool to narrow down datasets, items, or tasks by specific criteria. Alongside basic filtering, use Smart Search for intuitive queries, or DQL for advanced exploration.

Function

A single operation inside a module that can be executed when a service is deployed.

Group

Created within an organization, groups are reusable teams of users that can be assigned roles collectively in workflows and projects.

Integration/Secrets

Secure connections that store encrypted credentials (tokens, keys, passwords) for accessing cloud resources like GCS, S3, STS, and registries.

Item

A single file (image, video, text, audio, etc.) in a dataset. Each item can be annotated, labeled, or processed.

Item Status

A label showing progress or outcome of an item (e.g., Completed, Approved, Discarded).

Labels

Categories or classes used to create annotations (e.g., “cat,” “car,” “tree”).

Master Dataset

The original dataset that stores the actual files.

Member (organization role)

A user with limited permissions: can create projects and view members but cannot manage others or delete the organization.

Modality

A link between items, used to show overlays (e.g., multi-sensor images), replacements, or previews.

Model

A trained machine learning algorithm used for predictions or data labeling.

Node

A single step in a pipeline, such as dataset management, function execution, or ML training.

Ontology

The structure of labels and attributes in a project, forming the building blocks for training models.

Organization

A workspace in Dataloop, made up of users who collaborate on projects, share data, and manage resources together.

Owner (organization role)

The creator of an organization with full control (create projects, manage members, delete organization). Cannot be removed.

Package

A collection of static code (modules, functions, schema) used to build models or deploy services, processed using application technology.

Pipeline

An automated workflow of connected nodes that process data using humans, code, and ML models.

Project

A high-level workspace within an organization, focused on a specific goal. It manages datasets, tasks, and annotations.

Project Owner (project role)

Has full control of a project: can manage datasets, assign contributors, export data, set workflows, and delete the project.

Recipe

A set of labeling rules and instructions linked with an ontology, defining how data should be processed or annotated.

Service

A serverless computing service for running code without managing infrastructure. Conceptually, a deployed package serving the code.

Storage Driver

Bucket-specific connectors attaching external storage (e.g., GCS, S3, databases) to Dataloop, created from a storage integration.

Task

A unit of work (annotation, QA, review) assigned by project managers and containing assignments.

Trigger

A rule that starts an action automatically when conditions are met, either cron-based or event-based.

UI Slot

A button in the Dataloop UI that allows users to run functions directly within Annotation Studio, Dataset Browser, and other pages.

Worker (organization role)

A user role for annotators or workforce teams. They can work on tasks but have no access to secrets, projects, or organization settings. This role overrides project-level access.