Glossary

AI models

AI models can be Intel’s AI models, your models, or models from the Intel® Geti™ platform: Adaptive Training Sample Selection (ATSS), Single-Shot Detector (SSD), and YOLOX.

applications

Applications are single software products that will be deployed as part of a package on a host.

application package

An application package is a set of applications with specific versions deployed on a host.

artifacts

Any digital product generated during the training process of an AI model.This can include fully trained models, model checkpoints, intermediate files, or generated data.

bring your own container (BYOC)

BYOC is a custom container that you build and use.

bring your own source (BYOS)

BYOS is a custom code that you build and use.

building block

A building block is a sample application demonstrating a solution for a customer use case.

canvas

A canvas is a whiteboard for you to create and configure video pipelines.

Continuous Integration and Continuous Deployment (CI/CD)

A CI/CD pipeline consists of iterative and automated steps to build, test, and deliver a new software version.

cloud connector

A cloud connector connects items brought from a cloud platform (for example, Azure* and AWS*) to Edge Developer Toolbox.

cluster

A cluster is a group of edge nodes that run containerized, virtualized, or both applications.

containers

Containers are package types, which can be Intel’s or your own containers (See bring your own container (BYOC)).

customer

A customer is the end customer’s name for whom deployment is done.

dashboard

A dashboard lets you start, monitor, view performance information, and stop and remove your active project.

deployment

A deployment is an instance of a package configured to be installed on one or more hosts.

development environment

A high-code environment for you to work on.

Dockerfile

A text file that defines the steps to build a Docker image. This file contains instructions to install software, copy files, set environment variables, and set the entry point.

Docker Compose file

A text file in YAML format that defines and runs multi-container applications. This file usually describes multiple services (containers) and their configurations, including networking, volumes, ports, and dependencies.

File System

File System enables you to manage your configured projects, imported resources, and storage file system.

flow

A flow is a set of steps to get a specific output.

Intel® Geti™ software platform

The Intel® Geti™ software platform enables teams to build computer vision models for their AI applications.

GStreamer* framework

To quote from the GStreamer Framework website, “GStreamer is a framework for creating streaming media applications.”

Helm chart

A Helm* Chart is a package type. This chart contains all the resource definitions necessary to run an application, tool, or service inside a Kubernetes* cluster. A Helm Chart is like the Kubernetes equivalent of a Homebrew formula, an Advanced Packaging Tool (APT) dpkg, or a Yum RPM file. You can import Helm charts either from existing repositories like Bitnami* Application Catalog and Artifact Hub Application or a source code repository.

high-code

A high-code application is usually a complex application written using your chosen programming languages, tools, APIs, and SDKs.

host

A host is a single physical machine located on-premises at a customer site.

(to) launch

To launch a pipeline as a container means to run the pipeline on edge nodes for evaluation of results.

marketplace

A marketplace enables you to pre-build a container sample application to try out or use as building blocks for your projects.

media pipeline

A media pipeline of an application consists of basic building blocks that perform certain operations. The media pipeline transforms data into insights through audio and video processing, inference, and analytics operations. Each pipeline is fed by data sources such as a camera or RTSP streams, which go through data normalization operations such as decoding, resizing, color space conversion, cropping, resizing, and executing AI models.

metadata

Metadata has identifiable attributes that determine the packages deployed on specific hosts.

microservice

A microservice is a small independent service with well-defined interfaces and operations. It is independently buildable, testable, and deployable. It is used to build an application as a suite of small services, each running in its process. Microservices architectures make applications easy to scale and fast to develop, accelerating time-to-market for new features.

model creation

The Model Creation feature allows you to build AI models tailored to your desired task, such as object detection. You can create a new model by selecting the right dataset, annotating it, selecting a pre-trained model to retrain with new data, or training a model from scratch by identifying the right statistical or neural network and tuning the training parameters until a desired accuracy is reached.

model optimization

The Model Optimization feature enables you to optimize existing models from Open Model Zoo (OMZ), create new models, and import custom models.

module

A module is the main division of a package. A module is a stand-alone unit of software that you can choose to include in a package.

My Dashboard

The My Dashboard page provides a list of your saved applications, and a summary of status, execution time, and performance.

My Workspace

The My Workspace page enables you to edit and manage resources that you create.

node

An edge node is the Intel® platform on which your application runs.

package

A package is a set of software modules that are bundled together and prevalidated to ensure compatibility.

parameters

There are two types of parameters: hardware and software. Hardware parameters are related to performance monitoring. Hardware parameters include GPU power, GPU temperature, and GPU memory. Software parameters are input variables that can be configured. Software parameters are training parameters in an AI model or environment variables in test container configurations.

projects

One or more projects can be associated with imported resources, to be configured and run.

region

A region is an area that contains multiple sites. A region can also contain all sites in multiple states or cities. Examples: West, Northwest, and South.

registry

A registry is a place to create, store, secure, and manage containers.

resources

Resources are assets that can be imported. Examples: container images, Dockerfiles, Helm charts, Docker* Compose files, and application source code.

sample application

A sample application is pre-configured but can be customized or configured.

site

A site is the physical location where the customer’s host is deployed, and where configuration, debugging, and analysis may occur. Examples include restaurants and gas stations. A site is a starting point for identifying deployment details and troubleshooting.

storage

Storage can be either Intel’s or your own storage (See Bring Your Own Storage).

use case

A use case is a specific situation in which a product or service could be used to solve a problem.