Inspiration
The inspiration for this project came from observing the frequent misplacement of waste in recycling bins at our university. Improper waste sorting reduces recycling efficiency, contaminates recyclable materials, and increases landfill waste. We wanted to leverage AI technology to solve this issue and contribute to environmental sustainability by improving the accuracy of waste disposal in public spaces.
What it does
The AI-powered waste sorting system uses computer vision to recognize different types of waste—such as cardboard, glass, metal, paper, plastic, and trash—by scanning the items. It then suggests the appropriate recycling bin for the user or automatically sorts the waste into the correct bin when integrated into automated systems. This improves recycling accuracy and efficiency, reducing contamination and waste management costs.
How we built it
We trained a Convolutional Neural Network (CNN) using a dataset of 6,790 labeled images spread across six categories. The system was built using PyTorch for model development, AWS S3 for data storage, and AWS SageMaker Studio to speed up the training process with cloud-based compute resources.
Challenges we ran into
One of the challenges we faced was processing and normalizing large datasets to ensure the model could effectively classify different types of waste. Additionally, learning AWS tools like SageMaker and S3 was a new experience for the team, and integrating them into the project required extra effort. Lastly, ensuring that the model generalizes well to various waste items beyond those in the training data proved challenging.
Accomplishments that we're proud of
We successfully trained a CNN that can accurately classify waste into six categories with high precision. Additionally, we managed to scale our training process using AWS infrastructure, which significantly reduced training time and enhanced performance. We’re proud that we have built a solution with potential real-world applications in improving recycling efficiency and waste management practices.
What we learned
We gained hands-on experience with AI, machine learning, and cloud-based tools, including PyTorch and AWS SageMaker. We also learned about the challenges of training machine learning models on real-world data and how to optimize them for better performance. This project helped us improve our problem-solving and collaboration skills, as we worked together to build a meaningful solution.
What's next for AI-Powered Waste Sorting and Recycling System
The next step for this project is to refine the model for better accuracy and deploy the system in real-world environments, such as public spaces or university campuses. We aim to integrate the system with automated sorting machines to further streamline the recycling process. Additionally, expanding the dataset to include more waste categories and testing the system in various settings are key priorities for future development.
Log in or sign up for Devpost to join the conversation.