sustAInably

Inspiration for sustAInably

Recycling contamination is a major issue, with misidentified waste rendering entire batches unrecyclable. Existing solutions lack real-world accuracy, making it difficult for users to sort properly. Therefore, I wanted to create an AI-powered tool that simplifies waste sorting and effortlessly empowers individuals to make more sustainable choices.

What sustAInably does

sustAInably uses deep learning to classify plastic waste based on resin codes from images, guiding users on proper disposal. By reducing sorting errors, it helps prevent contamination in recycling streams and ensures waste is processed correctly.

How I built sustAInably

I trained a deep learning model on a dataset of real-world plastic waste images, refining it for accuracy in recognizing resin codes. Using TensorFlow, I developed a user-friendly interface that allows users to upload images and receive instant classification results.

Challenges I ran into

One challenge was data imbalance: certain plastic types, like plastic resin class 5, were overrepresented, skewing the model’s predictions. Initially, I tried weighting the classes and removing excess data, but this led to new biases. After multiple iterations, I refined the dataset and adjusted the training process to achieve balanced and accurate classification.

Accomplishments that I'm proud of

I built a functional AI model that accurately classifies plastic waste, making recycling more accessible. My model’s real-world accuracy and ability to distinguish tricky materials set it apart from existing solutions.

What I learned from this experience

I gained hands-on experience in training deep learning models with imbalanced data and learned the importance of refining datasets for real-world performance. Additionally, I saw firsthand how AI can drive sustainability efforts when combined with thoughtful design.

What's next for sustAInably

I plan to expand my dataset, improve classification for mixed-material waste, and integrate the tool into a mobile app for wider accessibility. Future updates may include multilingual support and collaboration with waste management organizations.