Welcome to Jetson Orin NANO Developer Kit repository
05.Pose estimation
Pose Estimation
Pose Estimation1. Model Introduction2. Pose Estimation: ImageResult Preview3. Pose Estimation: VideoResult Preview4. Pose Estimation: Real-Time Detection4.1. USB CameraResult Preview4.2. CSI CameraResult PreviewReferences
Uses Python to demonstrate Ultralytics Pose Estimation on images, videos, and real-time detection.
1. Model Introduction
Pose estimation is a task that involves identifying the location of specific points in an image, usually referred to as keypoints. The keypoints can represent various parts of the object such as joints, landmarks, or other distinctive features. The locations of the keypoints are usually represented by a set of 2D [x, y] or 3D [x, y, visible] coordinates
The output of a pose estimation model is a set of points that represent the keypoints of the objects in the image, and usually also the confidence scores for each point. Pose estimation is a good choice when you need to identify specific parts of the objects in a scene and their positions relative to one another.
The default pose model of YOLO26 has 17 keypoints, each representing a different part of the human body. Below is the mapping of each index to the corresponding body joint:
Nose
Left eye
Right eye
Left ear
Right ear
Left shoulder
Right shoulder
Left elbow
Right elbow
Left wrist
Right wrist
Left hip
Right hip
Left knee
Right knee
Left ankle
Right ankle
2. Pose Estimation: Image
Use yolo26n-pose.engine to predict the images in the ultralytics26 folder (not the images bundled with ultralytics).
Enter the code folder:
cd /home/jetson/ultralytics_yolo26/yahboom_demo
Run the code:
xxxxxxxxxxpython3 03.pose_image.py
Result Preview

Example code:
ximport os, syssys.path.insert(0, "/home/jetson/ultralytics_yolo26/runtime")os.environ.setdefault("ULTRALYTICS_CONFIG_DIR", "/home/jetson/ultralytics_yolo26/.config")
from ultralytics import YOLO
# Load a modelmodel = YOLO("/home/jetson/ultralytics_yolo26/models/yolo26n-pose.engine") # pretrained yolo26n-pose model
# Run batched inference on a list of imagesresults = model("/home/jetson/ultralytics_yolo26/assets/people.jpg") # return a list of Results objects
# Process results listfor 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/jetson/ultralytics_yolo26/output/people_output.jpg") # save to disk3. Pose Estimation: Video
Use yolo26n-pose.engine to predict the video in the ultralytics26 folder (not the video bundled with ultralytics).
Enter the code folder:
xxxxxxxxxxcd /home/jetson/ultralytics_yolo26/yahboom_demo
Run the code:
xxxxxxxxxxpython3 03.pose_video.py
Result Preview
The location of the video output by yolo recognition: /home/jetson/ultralytics_yolo26/output/

Example code:
xxxxxxxxxximport os, syssys.path.insert(0, "/home/jetson/ultralytics_yolo26/runtime")os.environ.setdefault("ULTRALYTICS_CONFIG_DIR", "/home/jetson/ultralytics_yolo26/.config")
import cv2from ultralytics import YOLO
# Load the YOLO modelmodel = YOLO("/home/jetson/ultralytics_yolo26/models/yolo26n-pose.engine")
# Open the video filevideo_path = "/home/jetson/ultralytics_yolo26/videos/people.mp4"cap = cv2.VideoCapture(video_path)
# Get the video frame size and frame rateframe_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 videooutput_path = "/home/jetson/ultralytics_yolo26/output/03.people_animals_output.mp4"fourcc = cv2.VideoWriter_fourcc(*'mp4v') # You can use 'XVID' or 'mp4v' depending on your platformout = cv2.VideoWriter(output_path, fourcc, fps, (frame_width, frame_height))
# Loop through the video frameswhile 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 windowcap.release()out.release()cv2.destroyAllWindows()
4. Pose Estimation: Real-Time Detection
4.1. USB Camera
Use yolo26n-pose.engine to predict the USB camera feed.
Enter the code folder:
xxxxxxxxxxcd /home/jetson/ultralytics_yolo26/yahboom_demo
Run the code: click the preview window and press the q key to stop the program!
xxxxxxxxxxpython3 03.pose_camera_usb.py
Result Preview
The location of the video output by yolo recognition: /home/jetson/ultralytics_yolo26/output/

Example code:
xxxxxxxxxximport os, syssys.path.insert(0, "/home/jetson/ultralytics_yolo26/runtime")os.environ.setdefault("ULTRALYTICS_CONFIG_DIR", "/home/jetson/ultralytics_yolo26/.config")
import cv2from ultralytics import YOLO
# Load the YOLO modelmodel = YOLO("/home/jetson/ultralytics_yolo26/models/yolo26n-pose.engine")
# Open the cammeracap = 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 rateframe_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 videooutput_path = "/home/jetson/ultralytics_yolo26/output/03.pose_camera_usb.mp4"fourcc = cv2.VideoWriter_fourcc(*'mp4v') # You can use 'XVID' or 'mp4v' depending on your platformout = cv2.VideoWriter(output_path, fourcc, fps, (frame_width, frame_height))
# Loop through the video frameswhile 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 windowcap.release()out.release()cv2.destroyAllWindows()
4.2. CSI Camera
Use yolo26n-pose.engine to predict the CSI camera feed.
Enter the code folder:
xxxxxxxxxxcd /home/jetson/ultralytics_yolo26/yahboom_demo
Run the code: click the preview window and press the q key to stop the program!
xxxxxxxxxxpython3 03.pose_camera_csi.py
Result Preview
The location of the video output by yolo recognition: /home/jetson/ultralytics_yolo26/output/

Example code:
xxxxxxxxxximport os, syssys.path.insert(0, "/home/jetson/ultralytics_yolo26/runtime")os.environ.setdefault("ULTRALYTICS_CONFIG_DIR", "/home/jetson/ultralytics_yolo26/.config")
import cv2from ultralytics import YOLOfrom jetcam.csi_camera import CSICamera
# Load the YOLO modelmodel = YOLO("/home/jetson/ultralytics_yolo26/models/yolo26n-pose.engine")
# Open the camera (CSI Camera)cap = CSICamera(capture_device=0, width=640, height=480)
# Get the video frame size and frame rateframe_width = 640frame_height = 480fps = 30
# Define the codec and create a VideoWriter object to output the processed videooutput_path = "/home/jetson/ultralytics_yolo26/output/03.pose_camera_csi.mp4"fourcc = cv2.VideoWriter_fourcc(*'mp4v') # You can use 'XVID' or 'mp4v' depending on your platformout = cv2.VideoWriter(output_path, fourcc, fps, (frame_width, frame_height))
# Loop through the video frameswhile True: # Read a frame from the camera frame = cap.read()
if frame is not None: # 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 no frame is received (camera error or end of stream) print("No frame received, breaking the loop.") break
# Release the video capture and writer objects, and close the display windowcap.release()out.release()cv2.destroyAllWindows()