YOLO26-X-pose
Ultralytics YOLO26 is the latest evolution in the YOLO series, engineered from the ground up for edge and low-power devices. This is the pose estimation variant optimized for human keypoint detection.
Model Specifications
| Property | Value |
|---|
| Input Size | 640 pixels |
| mAP Pose (50-95, e2e) | 71.6 |
| mAP Pose (50, e2e) | 91.6 |
| CPU Speed (ONNX) | 565.4 ms |
| T4 TensorRT10 Speed | 12.2 ms |
| Parameters | 57.6M |
| FLOPs | 201.7B |
Key Features
The architecture of YOLO26 is guided by three core principles:
Simplicity: YOLO26 is a native end-to-end model, producing predictions directly without the need for non-maximum suppression (NMS). By eliminating this post-processing step, inference becomes faster, lighter, and easier to deploy in real-world systems.
Deployment Efficiency: The end-to-end design cuts out an entire stage of the pipeline, dramatically simplifying integration, reducing latency, and making deployment more robust across diverse environments.
Training Innovation: YOLO26 introduces the MuSGD optimizer, a hybrid of SGD and Muon — inspired by Moonshot AI's Kimi K2 breakthroughs in LLM training. This optimizer brings enhanced stability and faster convergence.
Precision Pose Estimation Features
- Residual Log-Likelihood Estimation (RLE): More accurate keypoint localization
- Optimized Decoding: Increased inference speed for pose estimation
- DFL Removal: Simplified inference and broader hardware compatibility
- Up to 43% Faster CPU Inference: Optimized for edge computing
Usage
Install ultralytics with pip install ultralytics.
Download the model.
1from huggingface_hub import hf_hub_download
2
3model_path = hf_hub_download(repo_id="openvision/yolo26-x-pose", filename="model.pt")
Infer.
1from ultralytics import YOLO
2from PIL import Image
3import requests
4
5model = YOLO(model_path)
6
7url = 'http://images.cocodataset.org/val2017/000000039769.jpg'
8image = Image.open(requests.get(url, stream=True).raw)
9
10# Run inference with the YOLO26x-pose model on the image
11results = model.predict(image)
Documentation
For more information, see the
official YOLO26 documentation.
