WasteWizard is an AI consumer app that aims to tackle the critical issue of waste mismanagement by leveraging computer vision technology for automated waste classification and sorting. With the US recycling rate at just 32% (EPA 2018), optimizing waste sorting is urgent. Ultimately, this innovation promises to empower users with clear waste sorting guidance, enhance recycling efficiency, decrease landfill overflow, and mitigate environmental hazards associated with improper waste disposal.
A computer vision model that categorizes waste into 26 sub-categories of recyclables, e-waste, and hazardous materials
A UI for users to upload images and receive information on their appropriate waste categories and disposal methods
User analytics that describe recycling history, statistics, and tips
Facilitate an increase in the number of households and commercial properties who participate in waste sorting and improve accuracy of sorting.
73% of US households have recycling access but only 43% of households recycle
Make waste sorting machines more accurate and efficient, allowing workers to step away from manual sorting.
There are 300 MRFs in the US with over 9,000 workers perform manual sorting
Reduce the amount of waste going to landfills, and better the environment by prioritizing material reuse.
We send ~140 million tons of waste to landfills annually, but ~80% of items buried in landfills could have been recycled
WasteWizard's computer vision model involves a comprehensive collection of waste images sourced from TrashBox and Kaggle waste datasets. These images are utilized to train our model, enabling accurate categorization into 8 main waste categories and 26 subcategories. Our machine learning model predicts the presence of all classes and employ a final conditional statement to determine recyclability. This classification forms the basis for providing users with precise disposal methods.
Our user interface offers a seamless experience, allowing users to log in, upload photos, and view results effortlessly. With features like drag-and-drop functionality and image upload guidelines, users can ensure optimal model performance. Results are displayed on the Results page, complete with classification outcomes and recycling tips, while the Learning Center tab provides comprehensive guidance on waste disposal and management.
Our solution harnesses a variety of tools and libraries, including AWS services like EC2 and SageMaker, Python with PyTorch and TensorFlow, and essential packages like Scikit-learn and NumPy, ensuring comprehensive support for waste sorting AI application development. Through experimentation, we optimized model performance using techniques like class weight adjustment, regularization, and hyperparameter tuning, bolstered by transfer learning and ensembling methods for enhanced efficiency and accuracy.
Our architecture seamlessly integrates user interactions via UI, leveraging Amazon S3 for file storage and AWS Lambda for event-triggered EC2 instance execution, resulting in efficient model processing. The outcomes are stored in AWS DynamoDB via AWS AppSync, enabling quick access and delivery of results to the user's browser for interactive exploration.
.png)