Pose Detection Examples

from yolozone import PoseDetector
import cv2

# Initialize detector with a specific model
detector = PoseDetector(model="yolov8l-pose.pt")

# Read image
img = cv2.imread("exercise.jpg")

# Find keypoints
keypoints = detector.find_keypoints(img)

# Get points and lines for visualization
points, lines = detector.draw_pose(keypoints)

# Draw basic pose skeleton
for point in points:
    cv2.circle(img, point, 5, (255, 255, 255), 2)

for line in lines:
    cv2.line(img, line[0], line[1], (255, 255, 255), 2)

# Display or save result
cv2.imshow("Pose Detection", img)
cv2.waitKey(0)
cv2.destroyAllWindows()

def process_exercise_video(video_path, detector):
    """Analyze exercise form in a video"""
    cap = cv2.VideoCapture(video_path)
    
    # Get video properties
    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))
    
    while True:
        ret, frame = cap.read()
        if not ret:
            break
            
        # Detect pose keypoints
        keypoints = detector.find_keypoints(frame)
        
        # Draw basic skeleton
        points, lines = detector.draw_pose(keypoints)
        for point in points:
            cv2.circle(frame, point, 5, (255, 255, 255), 2)
        for line in lines:
            cv2.line(frame, line[0], line[1], (255, 255, 255), 2)
        
        # Analyze left arm angle (shoulder-elbow-wrist)
        angle, text, pos = detector.angle_between_3_points(
            keypoints, 5, 7, 9  # Left shoulder, elbow, wrist
        )
        cv2.putText(frame, f"L.Arm: {text}", pos, 
                   cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 255, 0), 2)
        
        # Measure hand distance (wrist to wrist)
        dist, text, pos, p1, p2 = detector.distance_between_2_points(
            keypoints, 9, 10  # Left to Right wrist
        )
        cv2.putText(frame, f"Hand dist: {text}", pos, 
                   cv2.FONT_HERSHEY_SIMPLEX, 0.8, (255, 0, 0), 2)
        cv2.line(frame, p1, p2, (255, 0, 0), 2)
        
        # Calculate arm symmetry (angle between arms)
        angle, text, pos = detector.angle_between_2_lines(
            keypoints, 5, 9, 6, 10  # Left arm vs Right arm
        )
        cv2.putText(frame, f"Arms angle: {text}", pos, 
                   cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 0, 255), 2)
        
        cv2.imshow("Exercise Analysis", frame)
        if cv2.waitKey(1) & 0xFF == ord('q'):
            break
    
    cap.release()
    cv2.destroyAllWindows()

# Initialize detector and process video
detector = PoseDetector(model="yolov8l-pose.pt")
process_exercise_video("exercise.mp4", detector)

# Initialize with larger model for better accuracy
detector = PoseDetector(model="yolov8x-pose.pt")

# Process video with detailed analysis
def analyze_exercise_form(video_path):
    cap = cv2.VideoCapture(video_path)
    frame_count = 0
    
    while True:
        ret, frame = cap.read()
        if not ret:
            break
            
        frame_count += 1
        
        try:
            # Detect pose
            keypoints = detector.find_keypoints(frame)
            
            # Draw skeleton
            points, lines = detector.draw_pose(keypoints)
            for point in points:
                cv2.circle(frame, point, 5, (255, 255, 255), 2)
            for line in lines:
                cv2.line(frame, line[0], line[1], (255, 255, 255), 2)
            
            # Analyze multiple angles
            angles = {
                'left_arm': detector.angle_between_3_points(
                    keypoints, 5, 7, 9)[0],  # Shoulder-elbow-wrist
                'right_arm': detector.angle_between_3_points(
                    keypoints, 6, 8, 10)[0],  # Shoulder-elbow-wrist
                'left_leg': detector.angle_between_3_points(
                    keypoints, 11, 13, 15)[0],  # Hip-knee-ankle
                'right_leg': detector.angle_between_3_points(
                    keypoints, 12, 14, 16)[0]  # Hip-knee-ankle
            }
            
            # Check form symmetry
            arm_diff = abs(angles['left_arm'] - angles['right_arm'])
            leg_diff = abs(angles['left_leg'] - angles['right_leg'])
            
            # Display analysis
            y_pos = 30
            for name, angle in angles.items():
                cv2.putText(frame, f"{name}: {angle:.1f}°", 
                          (10, y_pos), cv2.FONT_HERSHEY_SIMPLEX, 
                          0.6, (0, 255, 0), 2)
                y_pos += 25
            
            # Show symmetry analysis
            cv2.putText(frame, f"Arm symmetry diff: {arm_diff:.1f}°",
                      (10, y_pos), cv2.FONT_HERSHEY_SIMPLEX,
                      0.6, (0, 255, 255), 2)
            y_pos += 25
            cv2.putText(frame, f"Leg symmetry diff: {leg_diff:.1f}°",
                      (10, y_pos), cv2.FONT_HERSHEY_SIMPLEX,
                      0.6, (0, 255, 255), 2)
            
            cv2.imshow("Advanced Form Analysis", frame)
            if cv2.waitKey(1) & 0xFF == ord('q'):
                break
                
        except Exception as e:
            print(f"Error on frame {frame_count}: {str(e)}")
            continue
    
    cap.release()
    cv2.destroyAllWindows()

# Process video with advanced analysis
analyze_exercise_form("workout.mp4")