Background
Problem Statement
Project Objectives
Proposed Solution
System Workflow
Benefits
For Parents
For Healthcare Providers
For Researchers
For Society
Key Features
Expected Outcomes

Published May 30, 2026

Baby Positioning Detection System

Utilizing computer vision technology with several libraries from OpenCV to detect the baby's pose while in bed

BeginnerShowcase (no instructions)5 hours7

Things used in this project

Hardware components

Webcam, Logitech® HD Pro

Software apps and online services

Microsoft VS Code

Story

Background

Safe sleeping practices are critical for infant health and development, particularly during the first year of life. Medical organizations worldwide recommend that babies sleep on their backs (supine position) to reduce the risk of sleep-related incidents and support healthy breathing. However, continuous manual monitoring by parents or caregivers is often difficult, especially during nighttime or when caregivers are occupied with other activities.

With the advancement of Artificial Intelligence (AI), Computer Vision, and pose estimation technologies, it is now possible to automatically monitor a baby's sleeping position in real time using a camera-based system. By analyzing body key-points and posture patterns, the system can identify whether a baby is lying on their back, side, or stomach without requiring wearable sensors that may cause discomfort.

This project introduces an intelligent Baby Pose Detection System that leverages Computer Vision and machine learning techniques to provide continuous, non-contact monitoring of infant positions, helping caregivers ensure a safer sleeping environment.

Problem Statement

Parents and caregivers often face several challenges in monitoring infant sleeping positions:

Continuous supervision is difficult, particularly during nighttime or when performing daily activities.
Unintended position changes may occur while the baby is sleeping, increasing safety concerns.
Manual monitoring is time-consuming and may not provide immediate awareness of posture changes.
Existing monitoring solutions often rely on wearable devices or specialized sensors, which can be expensive, uncomfortable, or difficult to maintain.
Many households lack an affordable and intelligent system capable of automatically detecting infant posture in real time.

These challenges create a need for a reliable, automated, and non-invasive monitoring solution.

Project Objectives

The primary objective of this project is to develop a Computer Vision-based system capable of automatically detecting and classifying infant sleeping positions in real time.

Specific objectives include:

Detect the presence of a baby using a standard camera.
Extract body landmarks and key points through pose estimation algorithms.

Classify infant positions into categories such as:

Supine (Back Sleeping)
Prone (Stomach Sleeping)
Side Sleeping
Classify infant positions into categories such as:Supine (Back Sleeping)Prone (Stomach Sleeping)Side Sleeping
Display real-time position information and confidence scores.
Provide a foundation for future alert and notification systems when unsafe positions are detected.
Create a low-cost and scalable solution that can be deployed in homes, hospitals, and childcare facilities.

Proposed Solution

The Baby Pose Detection System utilizes a camera connected to an AI-powered processing unit. Video frames are analyzed continuously using Computer Vision algorithms and pose estimation models to identify key body landmarks, including the head, shoulders, arms, torso, hips, and legs.

The extracted skeletal information is then processed by a machine learning classification model that determines the baby's current position. The result is displayed in real time through an intuitive monitoring interface, showing both the detected position and prediction confidence.

Terlentang Position

Miring Position

System Workflow

Video acquisition from camera or recorded video.
Baby detection within the frame.
Body key point extraction using pose estimation.
Feature generation from skeletal coordinates.
Position classification using AI model.
Real-time visualization and monitoring dashboard.
Optional warning generation for unsafe sleeping positions.

Benefits

For Parents

Provides continuous monitoring without constant supervision.
Improves awareness of infant sleeping posture.
Reduces anxiety through real-time monitoring.
Supports safer sleeping practices.

For Healthcare Providers

Assists in monitoring infants in neonatal and pediatric care environments.
Enables objective posture tracking and analysis.
Supports data-driven observation of infant behavior.

For Researchers

Creates a platform for studying infant movement patterns.
Provides data for future AI and healthcare research.
Supports development of advanced infant monitoring systems.

For Society

Promotes awareness of safe infant sleeping positions.
Encourages adoption of AI-based healthcare technologies.
Contributes to improving infant safety and well-being.

Key Features

Real-time baby pose detection
Computer Vision-based monitoring
AI-powered posture classification
Contactless and non-invasive operation
High detection accuracy
Support for live camera and recorded video input
Expandable notification and alert system
Lightweight and cost-effective implementation

Expected Outcomes

The final outcome of this project is an intelligent monitoring system capable of accurately identifying infant sleeping positions in real time using Computer Vision technology. The system provides caregivers with actionable information regarding a baby's posture, helping improve safety, reduce monitoring effort, and establish a foundation for future smart nursery and healthcare applications.

import cv2
import mediapipe as mp

INPUT_VIDEO = "bayi.mp4"
OUTPUT_VIDEO = "baby_position_output1.mp4"

mp_pose = mp.solutions.pose
mp_draw = mp.solutions.drawing_utils

pose = mp_pose.Pose(
    static_image_mode=False,
    min_detection_confidence=0.5,
    min_tracking_confidence=0.5
)

cap = cv2.VideoCapture(INPUT_VIDEO)

if not cap.isOpened():
    print("Video tidak ditemukan")
    exit()

fps = cap.get(cv2.CAP_PROP_FPS)
width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))

fourcc = cv2.VideoWriter_fourcc(*'mp4v')

out = cv2.VideoWriter(
    OUTPUT_VIDEO,
    fourcc,
    fps,
    (width, height)
)

while True:

    ret, frame = cap.read()

    if not ret:
        break

    rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
    results = pose.process(rgb)

    position = "Tengkurap"

    if results.pose_landmarks:

        lm = results.pose_landmarks.landmark

        nose = lm[mp_pose.PoseLandmark.NOSE]

        left_eye = lm[mp_pose.PoseLandmark.LEFT_EYE]
        right_eye = lm[mp_pose.PoseLandmark.RIGHT_EYE]

        left_shoulder = lm[mp_pose.PoseLandmark.LEFT_SHOULDER]
        right_shoulder = lm[mp_pose.PoseLandmark.RIGHT_SHOULDER]

        # Visibility wajah
        face_visibility = (
            nose.visibility 
            
        ) 

        shoulder_width = abs(
            left_shoulder.x - right_shoulder.x
        )

        # Klasifikasi sederhana
        if face_visibility > 0.999 and shoulder_width > 0.020:
            position = "Terlentang"

        elif face_visibility < 0:
            position = "TENGKURAP"

        else:
            position = "Miring"

        mp_draw.draw_landmarks(
            frame,
            results.pose_landmarks,
            mp_pose.POSE_CONNECTIONS
        )

    cv2.putText(
        frame,
        f"POSISI: {position} {face_visibility} {shoulder_width}",
        (20, 50),
        cv2.FONT_HERSHEY_SIMPLEX,
        1,
        (0, 255, 0),
        2
    )

    out.write(frame)

    cv2.imshow("Baby Position Detection", frame)

    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

cap.release()
out.release()
cv2.destroyAllWindows()

print("Selesai")
print("Output:", OUTPUT_VIDEO)

Credits

Hardi Kurnianto

19 projects • 18 followers

Master student at Intelligent Control and Systems Engineering Department

Baby Positioning Detection System