Welcome to GarbageFlow !!


The topic of climate change caught our team's eye as we acknowledged the multitude of climate change-related problems that our planet faces. Through brain-storming and the process of elimination, we approached the concept of scanning an item's barcode to determine where it can be disposed of. This was ideated because some members of our team felt that sometimes, seeing 4 different trash options can be time-consuming and confusing, and some solve this problem by with a "simple" picture. Soon after, we realized that a bar code isn't as good of an idea as scanning an object, regardless of whether it even has a barcode. and determining the optimum disposal choice.

What it does

GarbageFlow is a machine learning application that uses computer vision to classify images of garbage into different categories, such as plastic, glass, metal, cardboard, and more. The user can take a photo of a piece of garbage using their webcam, and GarbageFlow will classify it using its trained model.

How we built it

For our project GarbageFlow, we utilized a combination of technologies to achieve our goals.

Firstly, we used computer vision to classify different types of waste into their respective categories such as plastic, paper, glass, metal, and organic waste. For this, we leveraged the PyTorch deep learning framework and pre-trained ResNet models to train our own custom model using a dataset of waste images.

Secondly, we built a web application using Flask, a Python web framework, which allowed users to upload images of waste to our system. Our model then classifies the image and provides information about the type of waste and how to recycle it correctly.

Challenges we ran into

We ran into a plethora of challenges while working on this project. First of all, we're all lower year CS & CE majors who have had little exposure to Machine Learning during our time within university. This meant that tackling many elements within our project would often lead us to a significant amount of self-learning within the 14 hour allocated time. Secondly, tackling a project idea which involved computer vision fell outside of our collective scope of knowledge. Learning how to work with PyTorch, TensorFlow, and researching through other technologies such as ResNet took up a lot of our working time.

Accomplishments that we're proud of

One of the biggest accomplishments that we're proud of is successfully implementing a machine learning model for the first time. We're also proud of the fact that we were able to tackle a project idea that involved computer vision, despite it being outside of our collective scope of knowledge.

What we learned

We gained experience working with PyTorch, TensorFlow, and ResNet, and we also improved our skills in data preprocessing and model training.

What's next for GarbageFlow

We plan to integrate GarbageFlow with mobile, most likely through a mobile app, to make it more accessible to users on the go. We also envision it becoming an educational tool to teach people about the importance of proper waste disposal and recycling.

Built With

Share this project: