Skip to main content
The Enrich tab is where AI-driven enrichment of the dataset is configured and monitored. It exposes the model picker, the preview of model output, and the progress view for running enrichment jobs. For the end-to-end workflow, see Enrich Your Datasets with AI Models.

Layout Overview

The Enrich tab is organized into three areas.
RegionPurpose
Model PickerSelect one or more models to run against the dataset
Preview PanelInspect a sample of model output before committing to a full run
Progress ViewTrack the status of running enrichment jobs and review past jobs

Model Picker

The model picker lists every enrichment model available for the dataset. Each entry shows the model name, its task type, and a short description. Selecting a model highlights it, and Add Models includes additional models in the same run. Available models cover a range of enrichment tasks:
Task TypePurpose
Object detectionDraws bounding boxes and classifies each detected object. Feeds the Objects view in the Explore tab.
Image taggingAssigns labels to the entire image. Populates the Labels section of the Insights Panel.
Image and object captioningGenerates descriptive text that powers Caption Search.
Semantic search embeddingsPowers Semantic Search across the dataset. For the theory, see How Semantic Search Understands Language.
Face detectionDetects faces and facial landmarks.
For the full catalog, including task type, descriptions, and model dependencies, see Model Catalog.
Some models require pre-existing enrichments before they can run. For example, VL-Object-Captioner requires Object Detection to have been applied first, and semantic search models require captions or embeddings from a prior enrichment step.

Preview Panel

The preview panel runs the selected model(s) on a small sample of images and displays the output so a reviewer can inspect the results before committing to a full dataset run. A Generate Models Preview button produces the sample, and when multiple models are selected the panel allows switching between them to compare outputs. An Enrich Dataset button commits every selected model to a full run across all images and frames in the dataset. For the end-to-end procedure, see Enrich Your Datasets.

Progress View

The progress view tracks enrichment jobs from start to finish. Each job entry shows the models being run, a live progress indicator, and a timestamp. Completed jobs remain in the view as history. Enrichment runs do not block exploration. You can continue to search, filter, and select in the Explore tab while a job is in flight, and newly generated metadata becomes available in the Insights Panel and filter options once the job completes. For live status across every dataset in the workspace, use the Task Manager. Notifications for completed enrichment runs appear in the bell icon in the dataset header — see Monitoring and Alerts for the full notification system.

Enrich Your Datasets

How enrichment works and the end-to-end run flow

Model Catalog

Every available enrichment model, task type, and dependency

Task Manager

Monitor every running task, including enrichment jobs

Explore Tab

Where enriched metadata becomes filterable and searchable