# AI & Models

## Overview

The [AI & Models](https://hub.luxonis.com/ai) area in Luxonis Hub supports both model delivery and the data-to-deployment loop.
Convert an existing model, manage versioned artifacts in the Model Registry, or collect and annotate data to train, evaluate, and
improve a model in Hub AI.

### Quick Conversion

Convert a model for a target RVC platform without creating registry history.

[Quick Conversion](https://docs.luxonis.com/cloud/hubai/quick-conversion.md)

### Model Registry

Manage model variants, versions, instances, conversion, and deployment references.

[Model Registry](https://docs.luxonis.com/cloud/hubai/model-registry/concepts.md)

### Training

Prepare datasets, run training and evaluation, and improve models from new data.

[Training](https://docs.luxonis.com/cloud/hubai/training/concepts.md)

## Quick Conversion

When you need the fastest path from a raw model file to a deployable Luxonis native format, the [Quick
Conversion](https://hub.luxonis.com/ai/tools) interface is the most direct option. It enables rapid, one-off conversions without
creating a model entry or maintaining version history. Just upload your model, convert it immediately, and download a variant
optimized for your Luxonis device.

## Model Registry

For long-term workflows, collaboration, or managing model variants, the [Model Registry](https://hub.luxonis.com/ai/models) serves
as the core of this area of Hub. Here, you can browse a curated collection of public models categorized by task and ready for
deployment, or upload your own models to maintain versions, share them with your team or the community, and keep track of updates
over time.

Unlike Quick Conversion — which focuses on speed and simplicity — the Model Registry includes a more robust workflow for
conversion called [Detailed Conversion](https://docs.luxonis.com/cloud/hubai/model-registry/detailed-conversion.md). This workflow
offers advanced configuration, supports multiple variants of a model, and fits naturally into versioned model management — making
it ideal for production-ready and collaborative development.

## Training

The [Training](https://docs.luxonis.com/cloud/hubai/training/concepts.md) section connects data preparation to versioned model
outputs:

 * Datasets: store samples, tasks, annotations, splits, tags, health checks, and read-only versions.
 * Projects: define the trainable tasks and connect compatible datasets to one or more model architectures.
 * Training runs: update weights and publish model versions; evaluation runs measure an existing checkpoint on validation or test
   data.
 * Credits: pay for run compute, with estimates reserved and actual usage settled when work ends.
 * Jobs: expose the background work behind uploads, exports, versioning, training, pre-annotation, and conversion.

Start with [Create, Upload & Export](https://docs.luxonis.com/cloud/hubai/training/datasets/upload.md) when you already have data,
or review [Training concepts](https://docs.luxonis.com/cloud/hubai/training/concepts.md) for the resource model first.

## The Hub AI loop

Collect or upload samples, annotate and complete them, check dataset health, and freeze a dataset version for a training run.
Training publishes a model version that can be evaluated, converted for deployment, or used to pre-annotate the next data batch.
Deployed applications produce new Snaps, bringing the workflow back to data collection.

Pipeline nodes:
- Collect or upload
- Annotate and version
- Train and evaluate
- Model version
- Convert and deploy
- Production Snaps

Pipeline connections:
- Collect or upload -> Annotate and version
- Annotate and version -> Train and evaluate
- Train and evaluate -> Model version
- Model version -> Convert and deploy
- Convert and deploy -> Production Snaps
- Production Snaps -> Collect or upload

See [The Hub AI workflow](https://docs.luxonis.com/cloud/hubai/training/workflow.md) for the full lifecycle, decision points, and
links to each guide.