> For the complete documentation index, see [llms.txt](https://docs.unitlab.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.unitlab.ai/documentation/get-started/welcome-to-unitlab.md).

# What is Unitlab

What Unitlab does, which data it supports, and where to start for your role.

Unitlab is a data annotation platform. Your team brings in raw data, picks the files worth labeling, labels them by hand or with AI, reviews the work and exports versioned training data, all in one place.

![Projects page with a card for each project, showing a cover image, item counts and progress](https://292810646-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FGjVLUz4wthGkGlRKM6rM%2Fuploads%2FmJQmbgskrGV44gozlWuu%2Fprojects-list.webp?alt=media)

## What you can do

| Step           | What it means in Unitlab                                                                                                                                                                                                                                                                                                                                                 |
| -------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| **Collect**    | [Upload files](https://docs.unitlab.ai/documentation/data/data-upload.md) or link [cloud storage](https://docs.unitlab.ai/documentation/data/cloud-providers.md) (S3, Google Cloud, Azure). Bring in labels you already have.                                                                                                                                            |
| **Curate**     | [Filter, search and find similar files](https://docs.unitlab.ai/documentation/data/data-curation.md), drop duplicates and blurry frames, and save the best ones as a versioned dataset.                                                                                                                                                                                  |
| **Label**      | Draw boxes, polygons, masks, keypoints, 3D cuboids, time ranges and text spans in the [Annotation Workbench](https://docs.unitlab.ai/documentation/annotations/annotation-workbench.md).                                                                                                                                                                                 |
| **Auto-label** | Let AI do the first pass: [Auto select](https://docs.unitlab.ai/documentation/auto-labeling/detect-anything-sam-1-sam-3.md) one object, [prompt](https://docs.unitlab.ai/documentation/auto-labeling/prompt-labeling.md) for every object of a class, or run a [model on the whole project](https://docs.unitlab.ai/documentation/auto-labeling/batch-auto-labeling.md). |
| **Review**     | Send each item through annotate, review and [quality checks](https://docs.unitlab.ai/documentation/qa/qa-overview.md), with comments and issues for feedback.                                                                                                                                                                                                            |
| **Export**     | Freeze the reviewed labels as a [release](https://docs.unitlab.ai/documentation/releases/releases-overview.md) in COCO, YOLO, UUEF and other formats.                                                                                                                                                                                                                    |

## Example: training a robot to grasp

A robotics team records its arm picking objects on a workbench.

1. **Collect.** They upload the camera frames to **Assets**, one folder per robot cell.
2. **Curate.** They filter out blurry frames, keep a balanced mix of objects, and publish the result as a **dataset**.
3. **Auto-label.** A Model stage outlines every graspable object before people start.
4. **Label.** Annotators fix the outlines and mark the gripper.
5. **Review.** A reviewer accepts or rejects each item.
6. **Export.** The team downloads a COCO **release** and trains the grasp model.

The same flow works for medical scans, satellite images, documents and sensor data.

## Supported data

Images, video, audio, text, PDF documents, web pages, tabular CSV, sensor time series, medical scans (DICOM, NIfTI, NRRD), pathology slides, geospatial rasters (GeoTIFF and others) and LiDAR point clouds. One project can mix several types, and related files, such as four camera views of the same scene, can be labeled together as one task.

See [Upload data](https://docs.unitlab.ai/documentation/data/data-upload.md) for every file format.

## Where to start

| You are...            | Start with                                                                                                                                                                                                                                                                                       |
| --------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| **New to Unitlab**    | [Quickstart](https://docs.unitlab.ai/documentation/get-started/end-to-end-quickstart.md): create a project, label one image and export it.                                                                                                                                                       |
| **An annotator**      | [Navigate the platform](https://docs.unitlab.ai/documentation/get-started/platform-navigation.md), then the [Annotation Workbench](https://docs.unitlab.ai/documentation/annotations/annotation-workbench.md).                                                                                   |
| **A project manager** | [Key concepts](https://docs.unitlab.ai/documentation/get-started/the-unitlab-object-model.md), then [Create a project](https://docs.unitlab.ai/documentation/projects/create-and-configure-a-project.md) and [Workflows](https://docs.unitlab.ai/documentation/workflows/workflows-overview.md). |
| **An ML engineer**    | [Import existing annotations](https://docs.unitlab.ai/documentation/data/import-local-annotations.md), [Releases](https://docs.unitlab.ai/documentation/releases/releases-overview.md) and the [Python SDK](https://docs.unitlab.ai/api-sdk-cli).                                                |
| **A workspace admin** | [Members and invitations](https://docs.unitlab.ai/documentation/collaboration/members-and-invitations.md), [Roles](https://docs.unitlab.ai/documentation/collaboration/role-based-access.md) and [Security](https://docs.unitlab.ai/documentation/security/security-overview.md).                |

## Related

[Key concepts](https://docs.unitlab.ai/documentation/get-started/the-unitlab-object-model.md) · [Navigate the platform](https://docs.unitlab.ai/documentation/get-started/platform-navigation.md) · [Quickstart](https://docs.unitlab.ai/documentation/get-started/end-to-end-quickstart.md)
