> 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-models/rf-detr.md).

# RF-DETR

## RF-DETR Object Detection

RF-DETR is Roboflow's transformer-based real-time detection model. Run inference against COCO-pretrained object detection checkpoints through the [Serverless Cloud API](https://docs.roboflow.com/deployment/roboflow-cloud/serverless-api), or self-host using [Roboflow Inference](https://docs.roboflow.com/deployment/self-hosted/self-hosted).

### RF-DETR Object Detection API

The steps below run RF-DETR through the [Serverless Cloud API](https://docs.roboflow.com/deployment/roboflow-cloud/serverless-api) and visualize results with [supervision](https://supervision.roboflow.com/).

{% stepper %}
{% step %}
**Get your API Key**

Create a Roboflow account, find your key on the [Roboflow API settings page](https://app.roboflow.com/settings/api) and make it available to your shell:

```bash
export ROBOFLOW_API_KEY="your-key-here"
```

{% endstep %}

{% step %}
**Install the dependencies**

These two packages call the model and draw its results:

```bash
pip install -U inference-sdk supervision
```

{% endstep %}

{% step %}
**Run the model**

Run `rfdetr-small` on a sample image and annotate boxes and labels:

```python
import os
import cv2
import supervision as sv
from inference_sdk import InferenceHTTPClient

image_url = "https://media.roboflow.com/quickstart/traffic.jpg"
image = sv.load_image_from_url(image_url)

client = InferenceHTTPClient(
    api_url="https://serverless.roboflow.com",
    api_key=os.environ["ROBOFLOW_API_KEY"],
)
result = client.infer(image, model_id="rfdetr-small")

detections = sv.Detections.from_inference(result)

annotated = sv.BoxAnnotator().annotate(image.copy(), detections)
annotated = sv.LabelAnnotator().annotate(annotated, detections)

cv2.imwrite("traffic-annotated.png", annotated)
```

<figure><img src="/files/PkIMz0tOIpCoeGfCVawd" alt=""><figcaption></figcaption></figure>
{% endstep %}
{% endstepper %}

{% hint style="info" %}
Set `api_url` to match your deployment target:

* `https://serverless.roboflow.com` for the Serverless Cloud API.
* `http://localhost:9001` for a local [Inference](https://docs.roboflow.com/deployment/self-hosted/self-hosted) server.
* Your [Dedicated Deployment](https://docs.roboflow.com/deployment/roboflow-cloud/dedicated-deployments) URL for a private endpoint.
  {% endhint %}

### RF-DETR Object Detection models and benchmarks

Pass any of these aliases as the `model_id` when running inference. The SDK resolves each alias to the underlying Roboflow project version.

<table data-search="false"><thead><tr><th>Alias</th><th>Input Size</th><th>mAP50-95</th><th>ONNX latency (ms)*</th><th>TensorRT FP16 (ms)*</th></tr></thead><tbody><tr><td><code>rfdetr-nano</code></td><td>384x384</td><td>48.4</td><td>9.7</td><td>6.2</td></tr><tr><td><code>rfdetr-small</code></td><td>512x512</td><td>53.0</td><td>12.9</td><td>8.3</td></tr><tr><td><code>rfdetr-medium</code></td><td>576x576</td><td>54.7</td><td>16.3</td><td>9.5</td></tr><tr><td><code>rfdetr-large</code></td><td>704x704</td><td>56.5</td><td>25.6</td><td>11.6</td></tr><tr><td><code>rfdetr-xlarge</code></td><td>700x700</td><td>58.6</td><td>41.6</td><td>14.9</td></tr><tr><td><code>rfdetr-2xlarge</code></td><td>880x880</td><td>60.1</td><td>53.4</td><td>21.7</td></tr></tbody></table>

## RF-DETR Instance Segmentation

RF-DETR also provides instance segmentation checkpoints that predict masks alongside boxes. Run them through the [Serverless Cloud API](https://docs.roboflow.com/deployment/roboflow-cloud/serverless-api), or self-host using [Roboflow Inference](https://docs.roboflow.com/deployment/self-hosted/self-hosted).

### RF-DETR Instance Segmentation API

Set your API key and install the dependencies as shown above, then run a segmentation checkpoint and draw its masks:

```python
import os
import cv2
import supervision as sv
from inference_sdk import InferenceHTTPClient

image_url = "https://media.roboflow.com/quickstart/traffic.jpg"
image = sv.load_image_from_url(image_url)

client = InferenceHTTPClient(
    api_url="https://serverless.roboflow.com",
    api_key=os.environ["ROBOFLOW_API_KEY"],
)
result = client.infer(image, model_id="rfdetr-seg-preview")

detections = sv.Detections.from_inference(result)

annotated = sv.MaskAnnotator().annotate(image.copy(), detections)
annotated = sv.LabelAnnotator().annotate(annotated, detections)

cv2.imwrite("traffic-annotated.png", annotated)
```

<figure><img src="/files/rkcpIH9tLlfqbUggpwox" alt=""><figcaption></figcaption></figure>

### RF-DETR Instance Segmentation models and benchmarks

Pass any of these aliases as the `model_id` when running inference. The SDK resolves each alias to the underlying Roboflow project version. Figures are mask mAP on COCO `val`.

<table data-search="false"><thead><tr><th>Alias</th><th>Input Size</th><th>Mask mAP50-95</th><th>ONNX latency (ms)*</th><th>TensorRT FP16 (ms)*</th></tr></thead><tbody><tr><td><code>rfdetr-seg-nano</code></td><td>312x312</td><td>40.3</td><td>15.8</td><td>10.5</td></tr><tr><td><code>rfdetr-seg-small</code></td><td>384x384</td><td>43.1</td><td>19.3</td><td>11.9</td></tr><tr><td><code>rfdetr-seg-medium</code></td><td>432x432</td><td>45.3</td><td>23.9</td><td>14.1</td></tr><tr><td><code>rfdetr-seg-large</code></td><td>504x504</td><td>47.1</td><td>30.1</td><td>15.4</td></tr><tr><td><code>rfdetr-seg-xlarge</code></td><td>624x624</td><td>48.8</td><td>51.2</td><td>19.1</td></tr><tr><td><code>rfdetr-seg-2xlarge</code></td><td>768x768</td><td>49.9</td><td>89.6</td><td>25.6</td></tr></tbody></table>

## RF-DETR Keypoint Detection

RF-DETR keypoint detection is a preview checkpoint, pretrained on COCO person keypoints. Run it through the [Serverless Cloud API](https://docs.roboflow.com/deployment/roboflow-cloud/serverless-api), or self-host using [Roboflow Inference](https://docs.roboflow.com/deployment/self-hosted/self-hosted).

{% hint style="info" %}
This checkpoint is a preview. Its accuracy and output format can change in later releases.
{% endhint %}

### RF-DETR Keypoint Detection API

Set your API key and install the dependencies as shown above, then run `rfdetr-keypoint-preview` and annotate keypoints:

```python
import os
import cv2
import numpy as np
import supervision as sv
from inference_sdk import InferenceHTTPClient

COCO_KEYPOINTS = [
    "nose", "left_eye", "right_eye", "left_ear", "right_ear",
    "left_shoulder", "right_shoulder", "left_elbow", "right_elbow",
    "left_wrist", "right_wrist", "left_hip", "right_hip",
    "left_knee", "right_knee", "left_ankle", "right_ankle",
]

image_url = "https://media.roboflow.com/notebooks/examples/person-walking.png"
image = sv.load_image_from_url(image_url)

client = InferenceHTTPClient(
    api_url="https://serverless.roboflow.com",
    api_key=os.environ["ROBOFLOW_API_KEY"],
)
result = client.infer(image, model_id="rfdetr-keypoint-preview")

# Each prediction lists only the keypoints the model can see, so place them in
# the 17 COCO slots the skeleton expects and mark the missing ones not visible.
xy, visible = [], []
for prediction in result["predictions"]:
    found = {point["class"]: (point["x"], point["y"]) for point in prediction["keypoints"]}
    xy.append([found.get(name, (0, 0)) for name in COCO_KEYPOINTS])
    visible.append([name in found for name in COCO_KEYPOINTS])

key_points = sv.KeyPoints(xy=np.array(xy, dtype=np.float32), visible=np.array(visible))

annotated = sv.EdgeAnnotator(color=sv.Color.GREEN, thickness=3).annotate(image.copy(), key_points)
annotated = sv.VertexAnnotator(color=sv.Color.RED, radius=5).annotate(annotated, key_points)
cv2.imwrite("person-walking-annotated.png", annotated)
```

<figure><img src="/files/8sEmE1TKvtBRLIQJjyZs" alt=""><figcaption></figcaption></figure>

### RF-DETR Keypoint Detection models and benchmarks

Pass the alias as the `model_id` when running inference. Accuracy is COCO `val` AP50-95 scored with object keypoint similarity (OKS), the standard COCO keypoint metric, so it does not compare to the box and mask mAP above.

<table data-search="false"><thead><tr><th>Alias</th><th>Input Size</th><th>Keypoint AP50-95</th><th>Parameters (M)</th><th>Latency (ms)†</th></tr></thead><tbody><tr><td><code>rfdetr-keypoint-preview</code></td><td>576x576</td><td>71.8</td><td>126.4</td><td>9.7</td></tr></tbody></table>

Roboflow does not publish a prebuilt TensorRT engine for this checkpoint yet, so Inference runs it on ONNX Runtime even when you install the `inference-models[trt10]` extra. Reaching the latency in the table means building the engine yourself with the [rfdetr](https://github.com/roboflow/rf-detr) package.

***

† Keypoint accuracy and latency are the figures published in the [RF-DETR benchmarks](https://rfdetr.roboflow.com/latest/#keypoints): latency is TensorRT FP16 on 1x NVIDIA T4 at batch size 1, timing the model only. That is a different GPU and a different pipeline from the tables above, so the two sets of latencies do not compare directly.

\* Latency is measured with [Roboflow Inference](https://docs.roboflow.com/deployment/self-hosted/self-hosted) on 1x NVIDIA L4, batch size 1, mean of 1,000 inferences (100 warmup). The default `inference-gpu` install runs ONNX on the CUDA execution provider; adding the `inference-models[trt10]` extra selects a prebuilt TensorRT FP16 engine automatically. FP16 matches FP32 accuracy within 0.2 mAP on COCO `val2017`. Accuracy is the published COCO `val` spec (see the [RF-DETR announcement](https://blog.roboflow.com/rf-detr/)).