Pose Estimation

Demonstrating Ultralytics Pose Estimation effects on images, videos, and real-time detection using Python.

1. Model Introduction

Pose estimation is a task that involves identifying the location of specific points (often called keypoints) in an image. Keypoints can represent parts of an object, such as joints, landmarks, or other notable features. Keypoint locations are usually represented as a set of 2D [x, y] or 3D [x, y, visible] coordinates.

The output of a pose estimation model is a set of points representing the keypoints of objects in the image, typically including a confidence score for each point. Pose estimation is a good choice when you need to identify specific parts of objects in a scene and their positional relationships to each other.

In the default YOLO26 pose model, there are 17 keypoints, each representing a different part of the human body. The following is the mapping of each index to the corresponding body joint:

1. Nose
2. Left Eye
3. Right Eye
4. Left Ear
5. Right Ear
6. Left Shoulder
7. Right Shoulder
8. Left Elbow
9. Right Elbow
10. Left Wrist
11. Right Wrist
12. Left Hip
13. Right Hip
14. Left Knee
15. Right Knee
16. Left Ankle
17. Right Ankle

2. Pose Estimation: Image

Use the yolo26n-pose_ncnn_model to predict images in the ultralytics folder (non-ultralytics built-in images).

Navigate to the code folder:

cd /home/sunrise/ultralytics/ultralytics/yahboom_demo_26_26

Run the code:

xxxxxxxxxx
python3 03.pose_image.py

Preview

yolo recognition output image location: /home/sunrise/ultralytics/ultralytics/output/

Sample code:

​x
from ultralytics import YOLO
​
# Load a model
model = YOLO("/home/sunrise/ultralytics/ultralytics/yolo26n-pose_ncnn_model")
# model = YOLO("/home/sunrise/ultralytics/ultralytics/yolo26n-pose.pt")
​
# Run batched inference on a list of images
results = model("/home/sunrise/ultralytics/ultralytics/assets/people.jpg")  # return a list of Results objects
​
# Process results list
for result in results:
    # boxes = result.boxes  # Boxes object for bounding box outputs
    # masks = result.masks  # Masks object for segmentation masks outputs
    keypoints = result.keypoints  # Keypoints object for pose outputs
    # probs = result.probs  # Probs object for classification outputs
    # obb = result.obb  # Oriented boxes object for OBB outputs
    result.show()  # display to screen
    result.save(filename="/home/sunrise/ultralytics/ultralytics/output/people_output.jpg")  # save to disk
​

3. Pose Estimation: Video

Use the yolo26n-pose_ncnn_model to predict videos in the ultralytics26 folder (non-ultralytics built-in videos).

Navigate to the code folder:

xxxxxxxxxx
cd /home/sunrise/ultralytics/ultralytics/yahboom_demo_26

Run the code:

xxxxxxxxxx
python3 03.pose_video.py

Preview

yolo recognition output video location: /home/sunrise/ultralytics/ultralytics/output/

Sample code:

xxxxxxxxxx
import cv2
from ultralytics import YOLO
​
# Load the YOLO model
model = YOLO("/home/sunrise/ultralytics/ultralytics/yolo26n-pose_ncnn_model")
# model = YOLO("/home/sunrise/ultralytics/ultralytics/yolo26n-pose.pt")
​
# Open the video file
video_path = "/home/sunrise/ultralytics/ultralytics/videos/people.mp4"
cap = cv2.VideoCapture(video_path)
​
# Get the video frame size and frame rate
frame_width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
frame_height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
fps = int(cap.get(cv2.CAP_PROP_FPS))
​
# Define the codec and create a VideoWriter object to output the processed video
output_path = "/home/sunrise/ultralytics/ultralytics/output/03.people_output.mp4"
fourcc = cv2.VideoWriter_fourcc(*'mp4v')  # You can use 'XVID' or 'mp4v' depending on your platform
out = cv2.VideoWriter(output_path, fourcc, fps, (frame_width, frame_height))
​
# Loop through the video frames
while cap.isOpened():
    # Read a frame from the video
    success, frame = cap.read()
​
    if success:
        # Run YOLO inference on the frame
        results = model(frame)
​
        # Visualize the results on the frame
        annotated_frame = results[0].plot()
​
        # Write the annotated frame to the output video file
        out.write(annotated_frame)
​
        # Display the annotated frame
        cv2.imshow("YOLO Inference", cv2.resize(annotated_frame, (640, 480)))
​
        # Break the loop if 'q' is pressed
        if cv2.waitKey(1) & 0xFF == ord("q"):
            break
    else:
        # Break the loop if the end of the video is reached
        break
# Release the video capture and writer objects, and close the display window
cap.release()
out.release()
cv2.destroyAllWindows()
​

4. Pose Estimation: Real-time Detection

4.1. USB Camera

Use the yolo26n-pose_ncnn_model to predict USB camera feed.

Navigate to the code folder:

xxxxxxxxxx
cd /home/sunrise/ultralytics/ultralytics/yahboom_demo_26

Run the code: Click on the preview window, press the q key to terminate the program!

xxxxxxxxxx
python3 03.pose_camera_usb.py

Preview

yolo recognition output video location: /home/sunrise/ultralytics/ultralytics/output/

Sample code:

xxxxxxxxxx
import cv2
from ultralytics import YOLO
​
# Load the YOLO model
model = YOLO("/home/sunrise/ultralytics/ultralytics/yolo26n-pose_ncnn_model")
# model = YOLO("/home/sunrise/ultralytics/ultralytics/yolo26n-pose.pt")
​
​
# Open the cammera
cap = cv2.VideoCapture(0)
cap.set(6, cv2.VideoWriter.fourcc('M', 'J', 'P', 'G'))
cap.set(cv2.CAP_PROP_FRAME_WIDTH, 640)
cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 480)
​
# Get the video frame size and frame rate
frame_width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
frame_height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
fps = int(cap.get(cv2.CAP_PROP_FPS))
​
# Define the codec and create a VideoWriter object to output the processed video
output_path = "/home/sunrise/ultralytics/ultralytics/output/03.pose_camera_usb.mp4"
fourcc = cv2.VideoWriter_fourcc(*'mp4v')  # You can use 'XVID' or 'mp4v' depending on your platform
out = cv2.VideoWriter(output_path, fourcc, fps, (frame_width, frame_height))
​
# Loop through the video frames
while cap.isOpened():
    # Read a frame from the video
    success, frame = cap.read()
​
    if success:
        # Run YOLO inference on the frame
        results = model(frame)
​
        # Visualize the results on the frame
        annotated_frame = results[0].plot()
​
        # Write the annotated frame to the output video file
        out.write(annotated_frame)
​
        # Display the annotated frame
        cv2.imshow("YOLO Inference", cv2.resize(annotated_frame, (640, 480)))
​
        # Break the loop if 'q' is pressed
        if cv2.waitKey(1) & 0xFF == ord("q"):
            break
    else:
        # Break the loop if the end of the video is reached
        break
# Release the video capture and writer objects, and close the display window
cap.release()
out.release()
cv2.destroyAllWindows()
​

4.2. CSI Camera

Use the yolo26n-pose_ncnn_model to predict CSI camera feed.

Navigate to the code folder:

xxxxxxxxxx
cd /home/sunrise/ultralytics/ultralytics/yahboom_demo_26

Run the code: Click on the preview window, press the q key to terminate the program!

xxxxxxxxxx
python3 03.pose_camera_csi.py

When this prompt appears in the terminal, press 2, then Enter.

Preview

yolo recognition output video location: /home/sunrise/ultralytics/ultralytics/output/

Sample code:

xxxxxxxxxx
import cv2
from SunriseRobotLib import Mipi_Camera
​
from ultralytics import YOLO
​
width=640
height=480
g_camera = Mipi_Camera(width, height, debug=False)
​
# Load the YOLO model
model = YOLO("/home/sunrise/ultralytics/ultralytics/yolo26n-pose_ncnn_model")
# model = YOLO("/home/sunrise/ultralytics/ultralytics/yolo26n-pose.pt")
​
# Set up video output
output_path = "/home/sunrise/ultralytics/ultralytics/output/03.pose_camera_csi.mp4"
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
out = cv2.VideoWriter(output_path, fourcc, 30, (640, 480))
​
while True:
    # Capture frame-by-frame
    ret, frame = g_camera.read()
    frame  =  cv2.resize(frame,(640,480))
​
    # Run YOLO inference on the frame
    results = model(frame)
​
    # Visualize the results on the frame
    annotated_frame = results[0].plot()
​
​
    # Write the frame to the video file
    out.write(annotated_frame)
​
    # Display the resulting frame
    cv2.imshow("Camera", annotated_frame)
​
    # Break the loop if 'q' is pressed
    if cv2.waitKey(1) == ord("q"):
        break
​
# Release resources and close windows
g_camera.release()
out.release()
cv2.destroyAllWindows()
​

References

Pose Estimation - Ultralytics YOLO Documentation