Tech Stack

IDEA

Detection

I used VGG net of CNN algorithm. because VGG net has Maxpooling(Minimize Size) and 3x3 Kernel(Convolution with image , Using with ReLu function)

So I can learning my data so fast and detail

As you can see Loss get close to zero and Accuracy get close to 100%

It means Deep Learning is doing good

When baby closed his/her eyes passing 5 seconds 'sleep' turn to 1, note baby is sleeping and play with white noisy music

Using Haar Cascade Algorithm we can detect baby's face whether they fall down from bed or not

Crying

Using Pyaudio Library and FFT(Fast Fourier Transform) analyze Realtime voice to frequency

From journalism baby's crying frequency is almost 400Hz So, i set threshold 350Hz and when over 5 times Note baby is crying and play calm music

Temperature

When moisture is lower than 70% Relay(switching moudle) turns to High(5V) to operate humidifier

And then, Upload to Firebase Server

React Native just load Temperature & Moisture Data from Firebase in Realtime

App

Parent can see their baby Realtime streaming and they can adjust option Camera or Music , humidifier

And if there were problems , Message Notification changed