In this growing world where everything is transforming into automation starting from our home to our car why not change the garage to automation with some implementation of AI and IOT to it so that the driver don's have to get out of the car for openning the garage manually .
What it does
Nowadays, Garage security is very vulnerable given the amount of theft activity that has gotten an increment over the last couple of years. Through this Smart IOT Garage Door System, we will mitigate a lot of risks and effort that a person has to go through while trying to get his/her car in or out. The Smart Garage system can automatically open the Garage door by facial recognition and number plate reading. Our system consists of an app which has been made with MIT App Inventory and through which manually, the Garage door can be opened and closed and also, we can control the lighting system of the Garage. Another feature is also included in this app with which we can see the last image of the person whose was face was registered during the opening of the app. Also, when the smart garage system is activated, a unique link is generated every time with which we can see the live feed of the garage front through the camera and also through any CCTV cameras if connected
How we built it
We had used the IBM cloud as the brain of our system , Here we had created a Cloudant DB and Object Storage to store the image of the car number plate or the person face which has been recognised by the system to make the door open. Moreover the Watson IOT cloud had been made for controlling the door and appliances with an app , which had been further made through the node-red-flow and with the MIT-App inventory. Now each credentials have been stored in the python script and each of the function had been given a separate python script to run the system in the JETSON NANO as its computing device taking the input from. the raspberry pi V2.0 camera.
Challenges we ran into
There are many challenges that we faced while building the project. Connecting the IBM cloud with Jetson Nano was a tough challenge. Also, as it is a preliminary prototype, there are certain things we have to look out for. The device will also respond when any known image of any trained person or number plate is brought to it’s sight. We can overcome this issue by focusing more on the OCR (Optical Character Recognition) part and on the facial recognition part with some extra eye-blinking cascades, etc. Also, the live stream is only possible in the local network and the user and the system must be connected over the same network to get the live video feed from the door cam. Again, the facial recognition part is a bit slow. And, the accuracy of the number plate detection OCR plate is only 60% for now i.e. according to our tests, it can detect 6 out of the 10 number plates correctly. But we can again improve it by rigorous training of the ML model.
Accomplishments that we're proud of
We are proud that we had learnt this much of things in just small amount of time and also by the response we get from is system with just small and simple coding is quite cherishing.
What we learned
We had learned how to use IBM cloud and to connect with the python script by installing the necessary package in the jetson nano, to make the app in the mit app inventory and to make the node red flow for the same also we had also learn about how to use the Jetson nano with the cloud server and to control it respectively.
What's next for Smart Garage system
We have to work on the following for getting a better response from our project:- i) Work on the ML part (number-plate-recognition & face recognition) more to make it more accurate ii) Work on the live streaming of the camera feed to access remotely iii)To make the UI of the app more unique and attractive.