> For the complete documentation index, see [llms.txt](https://docs.roboflow.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.roboflow.com/models/supported-tasks/keypoint-detection.md).

# Keypoint Detection

Find objects and a fixed set of points on each one (ex: the joints of a person for pose estimation), then train and deploy your own keypoint detection model on Roboflow.

## What is keypoint detection?

Keypoint detection finds each object and a fixed set of points on it. For each object, the model returns a class label, a box, and the position and confidence of each keypoint (ex: the shoulders, elbows, and knees of a person). The points connect into a skeleton that you define for your project. Use it when the position or angle of parts matters, not just where the object is.

<figure><img src="https://3448603185-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fcmo9mhfIjYrvFFy1U7dk%2Fuploads%2Fgit-blob-22e471520dbbe91d8447c9c6316da1b10f5eda76%2Fimg-task-keypoint-detection-wide.webp?alt=media" alt="A worker sorting oranges, with keypoints and a skeleton placed on the head, shoulders, arms, and hands"><figcaption></figcaption></figure>

## Supported keypoint detection models

Roboflow supports these keypoint detection models. See the [Keypoint Detection models](https://docs.roboflow.com/models/supported-models.md#keypoint-detection) table for export and deployment support per model.

<table data-search="false"><thead><tr><th>Model</th><th>License</th><th data-type="checkbox">Trainable on Roboflow</th><th>Latest release</th></tr></thead><tbody><tr><td><a href="https://docs.roboflow.com/models/supported-models/rf-detr.md">RF-DETR Pose</a></td><td><a href="https://docs.roboflow.com/models/model-licenses.md#platform-model-license">PML 1.0</a> (commercial use included on Roboflow cloud)</td><td>true</td><td>2026</td></tr><tr><td><a href="https://docs.roboflow.com/models/supported-models/rf-detr.md">RF-DETR (Preview)</a></td><td><a href="https://docs.roboflow.com/models/model-licenses.md">Apache 2.0</a> (free for commercial use)</td><td>true</td><td>2026</td></tr><tr><td><a href="https://docs.roboflow.com/models/supported-models/roboflow-3.md">Roboflow 3.0</a></td><td><a href="https://roboflow.com/licensing">Roboflow Licensing</a></td><td>true</td><td>2023</td></tr><tr><td><a href="https://docs.roboflow.com/models/supported-models/yolo26.md">YOLO26</a></td><td><a href="https://docs.roboflow.com/models/model-licenses.md">AGPL-3.0</a> (commercial use needs a license, included on Roboflow cloud)</td><td>true</td><td>2026</td></tr><tr><td><a href="https://docs.roboflow.com/models/supported-models/yolo11.md">YOLO11</a></td><td><a href="https://docs.roboflow.com/models/model-licenses.md">AGPL-3.0</a> (commercial use needs a license, included on Roboflow cloud)</td><td>true</td><td>2024</td></tr></tbody></table>

## Use a pre-trained RF-DETR Pose model

To find people and their keypoints without training, add a [Keypoint Detection Model](https://docs.roboflow.com/workflows/blocks/blocks/run-a-model/keypoint-detection-model) block to a [Workflow](https://docs.roboflow.com/workflows). In the model selector, open the "Pre-Trained" tab, pick RF-DETR Pose, and choose a size for each stage. The model is trained on COCO: it detects all 80 COCO classes and returns keypoints for people only.

## Fine-tuning a keypoint detection model

{% stepper %}
{% step %}

### Create a keypoint detection project

Create a [Project](https://docs.roboflow.com/datasets/create-and-upload/create-a-project) and pick "Keypoint Detection" as the project type. Then [define a skeleton](https://docs.roboflow.com/datasets/annotate/annotate/edit-keypoint-skeletons): the points to label and how they connect.

<figure><img src="https://3448603185-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fcmo9mhfIjYrvFFy1U7dk%2Fuploads%2Fgit-blob-b15a071a629926c02c28c53e9ec2567cfa311a29%2Fcreate-project-keypoint-detection.webp?alt=media" alt="Create project form with Keypoint Detection selected as the project type and a preview of a keypoint detection example"><figcaption></figcaption></figure>
{% endstep %}

{% step %}

### Upload and annotate images

[Upload images](https://docs.roboflow.com/datasets/create-and-upload/adding-data), then open each image in the [annotation editor](https://docs.roboflow.com/datasets/annotate/annotate/annotation-tools), draw a box around each object, and [drag each keypoint](https://docs.roboflow.com/datasets/annotate/annotate/annotate-keypoints) of its skeleton into place. Mark points that are hidden as occluded. AI labeling tools do not support keypoint projects, so you place keypoints by hand.

<figure><img src="https://3448603185-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fcmo9mhfIjYrvFFy1U7dk%2Fuploads%2Fgit-blob-61be286baf2707d0de361d2f0bb4896aa78178a2%2Fannotate-keypoints.webp?alt=media" alt="Annotation editor with a person box and a skeleton of keypoints placed on a skateboarder&#x27;s joints"><figcaption></figcaption></figure>
{% endstep %}

{% step %}

### Generate a dataset version

[Generate a dataset version](https://docs.roboflow.com/datasets/versions/dataset-versions/create-a-dataset-version): a frozen snapshot of your images with a train, valid, and test split, plus any preprocessing and augmentation.
{% endstep %}

{% step %}

### Train a keypoint detection model

Click "Train" and pick an architecture and size. See [Train a Model](https://docs.roboflow.com/models/train/train-a-model.md) for every option.

<figure><img src="https://3448603185-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fcmo9mhfIjYrvFFy1U7dk%2Fuploads%2Fgit-blob-c18ee2f4d0003a4d5a2a83168d241bb76ab131bb%2Ftrain-architectures-keypoint-detection.webp?alt=media" alt="Select Architecture step on the Train page listing RF-DETR Pose (Recommended), RF-DETR Preview, YOLO26, Roboflow 3.0, and YOLOv11"><figcaption></figcaption></figure>
{% endstep %}
{% endstepper %}

## Deploy a keypoint detection model

After training, deploy your keypoint detection model with the [Serverless Cloud API](https://docs.roboflow.com/deployment/roboflow-cloud/serverless-api), a [Dedicated Deployment](https://docs.roboflow.com/deployment/roboflow-cloud/dedicated-deployments), or on your own hardware with [Roboflow Inference](https://docs.roboflow.com/deployment/self-hosted/self-hosted). See [Choosing a Deployment](https://docs.roboflow.com/deployment/choosing-a-deployment) to compare the options.
