RF-DETR — Pose Estimation AI Model
RF-DETR Keypoint finds all 17 COCO joints — nose, eyes, ears, shoulders, elbows, wrists, hips, knees, ankles — for every person in one pass, drawn as an OpenPose skeleton ready to condition a ControlNet.
What it does
RF-DETR Keypoint finds each person in a photograph and returns where their joints are — the 17 standard COCO keypoints, from nose and eyes through shoulders, elbows and wrists to hips, knees and ankles.
Each person comes back as a separate skeleton, and the output is in OpenPose format, which is the standard input for pose-guided generation such as ControlNet Pose.
When to use it
- Reading what pose someone is in, joint by joint
- Driving character animation or motion transfer from a skeleton
- Sports and fitness form — a golf swing, a yoga pose, a running gait
- Range of motion, symmetry and posture tracked over time
- Understanding where people are and how a scene is arranged
Details
- CategoryPose Estimation
- Year2026
- LicenseApache-2.0
Compliance & Provenance
- ProviderRoboflow (open) · Specialized
- EU AI Act RiskMinimal Risk
- Art. 50 TransparencyNot applicable
Inputs & Outputs
- ImageInputimage
Source image for human pose estimation
- ImageOutputimage
OpenPose-format skeleton render (black background) — connect directly to pose ControlNet as the control image
- TextOutputstring
Per-person bounding box and keypoint coordinates with confidence scores
Parameters
- thresholdfloatdefault 0.5range 0–1
Minimum confidence score to keep a detected person (higher = fewer, more confident detections).