Try the Model

Use the widget below to experiment with RF-DETR Segmentation. You can detect COCO classes such as people, vehicles, animals, household items.

Overview

RF-DETR (“Roboflow Detection Transformer”) is a real-time, transformer-based architecture designed to transfer well across a wide range of domains and dataset sizes. RF-DETR Segmentation extends the RF-DETR architecture with a segmentation head, enabling real-time instance segmentation alongside object detection.

RF-DETR Segmentation is built to balance segmentation quality and inference efficiency. Benchmarks on datasets such as Microsoft COCO show strong performance compared to prior DETR-based segmentation approaches, while maintaining low end-to-end latency on modern GPUs. RF-DETR is released as Apache 2.0 open source.

RF-DETR Seg expands the RF-DETR family with a segmentation head inspired by MaskDINO, optimized for efficiency with a ViT backbone. The result is precise object masks with state-of-the-art accuracy, redefining what’s possible in segmentation.

You can now train RF-DETR Segmentation models with any Instance Segmentation project in Roboflow. These models can be deployed on offline on your edge devices with Inference and Roboflow Workflows. You can also train models with the open source rfdetr Python package.

Here is example showing RF-DETR Segmentation running on images:

RF-DETR Segmentation License

RF-DETR Segmentation

is licensed under a

Apache 2.0

license.

Performance

Deploy a RF-DETR Segmentation API

You can use Roboflow Inference to deploy a

RF-DETR Segmentation

API on your hardware. You can deploy the model on CPU (i.e. Raspberry Pi, AI PCs) and GPU devices (i.e. NVIDIA Jetson, NVIDIA T4).

Below are instructions on how to deploy your own model API.


import os
import supervision as sv
from inference import get_model
from PIL import Image
from io import BytesIO
import requests

url = "https://media.roboflow.com/dog.jpeg"
image = Image.open(BytesIO(requests.get(url).content))

model = get_model("rfdetr-seg-preview")

predictions = model.infer(image, confidence=0.5)[0]

detections = sv.Detections.from_inference(predictions)

labels = [prediction.class_name for prediction in predictions.predictions]

annotated_image = image.copy()
annotated_image = sv.MaskAnnotator(color=sv.ColorPalette.ROBOFLOW).annotate(annotated_image, detections)
annotated_image = sv.LabelAnnotator(color=sv.ColorPalette.ROBOFLOW).annotate(annotated_image, detections, labels)

sv.plot_image(annotated_image)

Label Data Automatically with RF-DETR Segmentation

You can automatically label a dataset using RF-DETR Segmentation with help from Autodistill, an open source package for training computer vision models. You can label a folder of images automatically with only a few lines of code. Below, see our tutorials that demonstrate how to use RF-DETR Segmentation to train a computer vision model.

No items found.