The world is a beautiful place. And it's best that we keep it that way. Which is why we created Trashify. Trashify is a program that is in development that will help organize different types of trash and recyclables. It will tell the user what type of trash they see, by the simple act of taking a picture. Trashify uses a large data sample and a neural network to predict what type of trash it is and the AI improves as more people use it.
How we built it
We built it using Python and tested it out in Jupyter. We had to import numpy for the calculations, and we had to follow the rules of making a neural network. After the template was made, we imported the images and taught the neural network what to do.
Challenges we ran into
One challenge we ran into was the debugging process. There were many errors with the np.dot or dot product function in NumPy that made matrix-vector multiplication annoying to do. With our images, we figured out a way to fix this by using the same dimensions on the vector. Still, the image calculations won't work on GitHub, however, we made a video to show it.
Accomplishments that we're proud of
Some accomplishments we're proud of our building a successful neural network inside Python. It was really challenging getting it right, but we managed to persevere in the end. Not only that, but we're also proud of our teamwork with each other. Without that, we would have definitely not finished.
What we learned
We learned a lot of things. First, we learned a lot about the debugging process and the challenges of debugging. Not only that, but programming is a vast place with so many different aspects. This was a great learning experience for our whole team.
What's next for Trashify
Trashify has a very bright future. Without the time constraint, we are free to expand Trashify and its features from a basic neural network demo to an advanced application. Instead of being in the command line, it could be a powerful application if fully thought through. We are proud of what it is so far, and what it could become.
Log in or sign up for Devpost to join the conversation.