The amount of trash in the world has increased exponentially. Most of it ends up in landfills. Recycling is a method to combat this, but due to heavy contamination such as putting trash in the recycling bin, recycling has become economically unfeasible.
Our app detects what type of trash the user wants to throw away, identifies the material, and determines if it can be recycled.
Using a 34 layer neural network, we built an image classifier to train a model to recognize what trash to throw away and what trash to recycle.
Training the neural network was exceptionally difficult. Fine tuning the parameters, determining the learning rate, fixing numerous bugs, preventing overfitting, and balancing computer and time resources with accuracy, were all major challenges. Also, connecting the backend to the frontend using Javascript, Python, and HTML was exceptionally difficult.
The machine learning model was able to detect different types of trash with 92% accuracy and determine which disposal method was best. In addition, we were able to create a functional UI that was simple to use.
Perseverance, dedication, and embracing the unknown were the major themes of this project. Numerous challenges were faced, but we viewed them as opportunities and grew our technical abilities, as well as, our tenacity.
In the future, we will recognize specific types of plastic to know which can be recycled and which need to be thrown away, as well as, advise users to buy products that don't generate much waste.
Built With
- fastai
- html
- javascript
- linux
- python
- trash-library
Log in or sign up for Devpost to join the conversation.