K NearestNEighbor prediction
Sound sensor, buzzer Arduino
Inspiration: America is the world's largest waste producer with around 500 million tons waste produced every year. Out of which 90% of the waste ain't been recycled every year. Yes 90%. Why? Its' time we become more responsible towards the waste we generate, if we want our next generations to witness the same nature of the earth.
What it does : We used magnetometer sensor of android phone to determine the Tesla Values of metals like steel, Almunium and plastic. We collected data set and implemented machine learning to clarify the material present in the waste by the unpredicted sense value of magnetometer. The amount of the money refunded to the user account depends on the material classified for particular user. We also created a Web Application and hosted on cloud for User to check the recent and the credit history of the waste produced by him.
How we built it : We used KNN and SVM Machine learning classifier to classify the sensed values through magnetometer when placed against the bin containing all the waste. We are able to achieve 95~98 percentage accuracy through the data set of around 10,000 entries collected by the sensor. Based on the material , a particular amount has been calculated in python which is to be refunded to the user wallet. We deployed the user app in the cloud where user can access the recent credits as well as the credit history based on the recycle metal present in his waste.
Challenges we ran into: Firstly, we were expecting IR proximity sensor, or a single sensor, which could help us detect all the waste accurately, but it is after lot of research, we came to know, that due to variety of waste we are dealing, we might need combination of sensor, and a Machine Learning based system, to increase the accuracy of total segregation of the recyclable waste to calculate the amount we need to give the user the rebate about.
Accomplishments that we're proud of : We are proud to propose the innovative idea which promote the waste management by more use of recycle products.
We are proud of implementing the machine learning algorithm to classifying the metal with great accuracy even in the absence of open source database of the Tesla values of different metals. We are also proud to showcase this idea in IoT sector in such a time constraint environment, when most of us are experiencing a Hackathon for the first time