Yolo Pose Estimation and Skeleton
From wikiluntti
Introduction
Make a pose estimator and use it to make a moving skeleton.
Use Yolo from Ultralytics.
- Python 3.7+
- Yolo v11
- A CUDA-enabled GPU (optional but recommended for faster inference).
pip install ultralytics opencv-python numpy
Yolo
There are 17 keypoints. YOLOv11’s pose model outputs:
- (x, y) coordinates for each keypoint and
- confidence scores indicating the model’s certainty in each keypoint’s position.
Image detection
-
AN image before and after the code.
from ultralytics import YOLO
import matplotlib.pyplot as plt
import cv2
from PIL import Image
model = YOLO("yolo11n-pose.pt") # n, s, m, l, x versions available
results = model.predict(source="sample_image.jpg")
plt.figure(figsize=(10, 10))
plt.title('YOLOv11 Pose Results')
plt.axis('off')
plt.imshow(cv2.cvtColor(results[0].plot(), cv2.COLOR_BGR2RGB))
The results list includes results[0].keypoints.xy and results[0].keypoints.conf data. Printing that gives some general information about what is found and how fast, and a tensor vector which includes the position data.
