Predicting Risky Environment for Child Inside House using Deep Learning

Abstract

We place a high priority on making sure kids are safe. Yet in the modern world, there are a lot of threats to children's safety and challenges to achieving other safety goals. Parent can't always supervise their kids and leave them alone inside the house. So this is a big problem of our time as the small kids is unable to assess the risk due to any harmful activity or objects. The purpose of this study is to provide safety to the child by alerting the parent about the child's position and surroundings. This proposed model can be used to monitor child's movements. We proposed a children's video data set. The video data set contains every moment of the child in the form of frames, and 30 frames are created every second. Additionally, we train a baseline computer vision module that determines the level of dangerousness for a child using deep learning. The baseline consists of well-established methods like conditional and logistic regression for child's pose estimation, as well as cutting-edge deep learning image classifiers. We then integrate the stumbled-condition baseline into various learning platforms and evaluate performance in terms of accuracy and loss.