trashCloud_Sort

sample inputs for the segregator
segregation output from webcam

Inspiration:

Observation that people don't segregate trash or are just confused about it. As students we have had multiple instances of being in a rush leading us to not be able to spare that extra second to segregate our trash

What it does:

Help people segregate trash appropriately without any manual effort or thinking.

How we built it:

trained our own neural network to identify different types of trash and deployed it on a raspberry pi which communicates with google cloud to help us further train our neural network to improve the overall functionality of the program

Challenges we ran into:

interfacing the raspberry pi with our exported neural network and getting it to work with the camera. We also head troubles integrating google cloud in our system in a streamlined way.

Accomplishments that we're proud of:

Successfully trained a neural network and deployed it on a raspberry pi. Understanding how to use ONNX and TensorFlow to run trained models.

What we learned:

training on TensorFlow, running neural nets using ONNX. Using shell scripts on python and a lot about google cloud compute along with machine learning Also that at times changing paths is better than being stuck on the same lane

What's next for trashCloud_Sort:

improving on the neural network and adding hardware functionality to the detected objects to automatically sort them Creating a large enough database and self-training our network on cloud to improve accuracy.