The Toddler Guard
About Toddler Guard
First of all, you have to see it with your eyes
Toddler Guard is designed to create a safer environment for toddlers by leveraging advanced technology to identify potential dangers and alert parents in real-time. Our motivation for this project stems from the need to provide caregivers with a reliable and efficient tool to monitor their child's surroundings, ensuring their safety even when direct supervision is not possible.
This project was initiated to address the need for better safety measures for toddlers in homes. Toddlers, by their nature, are curious and often engage with their environment in unpredictable ways, which can expose them to numerous hazards. Our target is to address safety challenges that arise when toddlers are left without the caregiver's attention and can be exposed to dangerous items and situations.
Our system integrates various technologies such as camera input streaming, object detection using YOLO, danger analysis based on spatial metrics, and real-time alerting to notify caregivers of potential hazards. By doing so, we aim to prevent accidents and injuries that could occur due to toddlers' natural curiosity and unpredictable interactions with their environment.
Workflow
Our real-time monitoring system works as follows:
Camera Input Streaming
Toddler Guard connects to existing camera sources, continuously streaming video from the room where the toddler is present. This allows the system to integrate with an existing setup, with no need for additional hardware.
Object Detection
The system uses YOLOv8Pose for pose estimation to detect and analyze the toddler's movements.
YOLOv5 for object detection identifies and classifies objects in the room as safe or hazardous.
MiDaS for depth calculation calculates the depth of objects and child’s keypoints.
Danger Analyzing
The system employs a multi-approach method to assess danger levels.
Extracts keypoints to understand the toddler's movements.
Identifies the 2D metric distance between the toddler and hazardous objects.
Calculates a weighted average depth of the keypoints.
Compares it with the depth information of the hazard to determine the level of danger.
Alerting
Sends an alarm to the application installed on the caregiver’s phone when danger is detected.
A website for monitoring the supervised child in real-time.
Telegram bot allows every family member to connect and receive notifications with a frame where danger was detected attached.
Our Team
Tanya Fainstein
Tatiana.fainstein@mail.huji.ac.il
Eitan Stepanov
Eitan.stepanov@mail.huji.ac.il
Electrical engineering and computer science students in The Rachel and Selim Benin School of Computer Science and Engineering at 4th year.
Mentor:
Daniella Har Shalom: Daniellah@gmail.com