Glossary¶

Inference uses some terms (and terms of art) that may be unfamiliar to some readers. This page aims to disambiguate and clarify their meaning, specifically in the context of the Roboflow and Inference ecosystem.

API¶

An interface that allows software applications to communicate with each other. In Roboflow, the API provides access to workflows and blocks, enabling users to configure and execute inference pipelines programmatically.

Block¶

The fundamental unit of a Roboflow Workflow. Blocks perform specific tasks, such running model inference, performing logic, or interfacing with external services.

CLI¶

Command-Line Interface. A tool used to interact with Inference, such as starting the server, running benchmarks, or executing Workflows.

Client¶

A software application or library that interacts with the API. Our Python SDK is a client that abstracts the REST API into a more user-friendly interface.

Commercial License¶

A license permitting businesses to use models under terms suited for commercial applications, typically involving subscription plans or usage fees. While Inference uses a permissive Apache 2.0 license, some models have terms that may require downstream users to open source their own codebase. A commercial license removes that restriction.

Dedicated¶

A configuration where cloud compute resources for running Inference are allocated exclusively to a single user or organization, ensuring optimal performance and enabling additional functionality. Billed based on running time & usage.

Definition¶

The structured JSON description of a Workflow, including its sequence of Blocks, input parameters, response format, and any custom logic.

Dynamic Block¶

A type of Workflows Block using custom Python Code included in the JSON Definition of the Workflow allowing advanced customization at runtime.

Enterprise¶

A service tier for large organizations, offering enhanced capabilities, scalability, and support. Source code for enterprise functionality is included in the enterprise folder of the repo but may only be used in conjunction with an active Enterprise license.

Execution Engine¶

The backend system responsible for executing Workflows, managing the execution of Blocks, and optimizing resource allocation.

Fine-Tuned Model¶

A model that has been trained on a specific dataset for improved performance in a targeted application. For example, a "Scratch Detection" model tuned to find scratches on a specific automotive component.

Foundation Model¶

A general-purpose model that knows a lot about a lot and does not necessarily need to be fine-tuned on a specific dataset and can be used "zero-shot".

Kind¶

A categorization of data types used in Workflows. Defining the input and output Kinds of a Workflow Block allows the Execution Engine to validate, serialize, and optimize data and connections. Example Kinds are detection (representing a prediction from an object detection model), image (containing pixels and their metadata), and float_zero_to_one.

Inference Pipeline¶

An asynchronous interface for video streaming that handles efficiently consuming and routing video frames from a camera source and through a Workflow while maintaining state.

LMM¶

Large Multimodal Model. A Block type that processes multiple data types, such as images and text. Examples are Florence-2 and OpenAI's GPT-4o.

Managed¶

A deployment running in Roboflow's Cloud environment where the scaling and infrastructure are provided as a service. Contrast with "Self-Hosted" where the customer installs the software on their own infrastructure and is responsible for its setup and maintenance.

Metered¶

A billing approach based on usage metrics, such as the number of Workflow executions or hours of video processed. Models and Workflows that require an API Key to access Roboflow's Cloud Services are metered.

Model¶

A trained machine learning artifact used for inference tasks such as object detection or classification. Consists of an architecture (like ResNet-32) and trained weights.

Parameter¶

An input to a Workflow. Can be an image or data like strings, numbers, arrays, or objects. Used as inputs to Blocks.

Platform¶

The Roboflow ecosystem, which includes end-to-end tools for collecting and organizing data, annotating images, creating datasets, training models, building Workflows, and monitoring deployments.

Pre-Trained Model¶

A model architecture (like YOLOv11) loaded with weights that have been trained on a generic dataset (like Microsoft COCO). Contrast with a Fine-Tuned model that has been trained on a domain-specific dataset.

Public¶

A dataset, model, or Workflow that is accessible to all users within the Roboflow ecosystem. These can be found and distributed on Roboflow Universe.

Schema¶

The structured format of input and output data for a Workflow, defining the properties and types a downstream application needs to pass and parse to integrate with the Workflow.

SDK¶

Software Development Kit. A set of tools and libraries provided for integrating with the API using user-friendly abstractions.

Server¶

A device running Inference's HTTP interface, usually through Docker on port 9001.

Serverless¶

An execution model for Inference where resources scale on demand, without requiring manual resource management. Contrast with Dedicated. Billed solely based on usage.

Traditional CV¶

Blocks in Workflows that implement traditional computer vision techniques, such as filtering or edge detection, without relying on machine learning models.

Universe¶

A large collection of datasets, pre-trained models, and other resources shared publicly by other users and available for use within the Roboflow ecosystem.

Weights¶

The learned parameters of a machine learning model determining its predictions for a given input.

Workflow¶

A series of interconnected Blocks designed to process data and produce desired outcomes through a defined sequence of operations. Can be used for chaining models, maintaining state, performing custom logic, and integrating with external systems.

Workspace¶

A collaborative environment in Roboflow where users can create datasets, train models, and build Workflows. A Workspace is the container for data and arbiter of access via seats and API Keys.