KidsGuardian

The Project

KidsGuardian is a mobile application that monitors children at home alone and sends alerts to parents when their childrens are in potential risks of danger. This app uses object detection and machine learning to detect if a child is near the window or the door. The hardware part of this project uses a Raspberry Pi with a Pi camera. The frontend of the app was developed using Flutterflow, connected to Firebase for database. The system is programmed using the YOLOv8 deep learning model, achieving high accuracy in object detection.

View this project on YouTube

Team Comments

I was inspired to create this project because I read news about children falling from the window when they are home alone. My cousin’s child was only 2 years old, and she is worried about her child since she’s busy with work. I wanted to offer a solution to children’s safety problem to help out.

The door detection model met the problem of overfitting due to having a limited amount of training data. The door dataset originally consisted of two classifications, “door” and “doors” for singular and multiple doors. I figured that having two classes was unnecessary in the scenario of my project.

If budget allows, I want to obtain a depth-sensing camera, which detects an object's position in three dimensions. The Raspberry-Pi camera I’m using now only detects in two dimensions. A depth sensing camera will make the detection more accurate.