A Novel Multifaceted Deep Learning-Based Mobile Application for Accurate and Efficient Waste Classification and Increased Composting Engagement in Communities

Abstract

Solid food waste is slowly accumulating around the world in landfills. This waste is hazardous to human health and our environment when it decomposes as it leads to widespread release of greenhouse gasses such as CO2 and methane. Composting household waste actively can put wasted food to good use by creating arable soil and help mitigate the waste crisis that causes climate change. But currently, the general public finds it hard to manage and classify exactly what item is compostable and non-compostable. Furthermore, there is lack of motivation, incentive, and community support for composting waste actively. This paper proposes CompostAI, a novel deep learning-based mobile application targeted to make community participation in composting easy and socially engaging. This application uses a Xception convolutional neural network (CNN) model to classify waste into seven categories: compost, paper, cardboard, glass, metal, trash, plastic. The Xception model demonstrated optimal performance as it had the highest accuracy of 78.43% and F1 score of 81.22 out of the six CNN model trained, validated, and tested. CompostAI also has supplemental features that include allowing users to announce local sustainability events, find nearby composting centers, and learn new sustainable living techniques. CompostAI successfully makes community participation in composting easy and socially engaging, increasing composting rates and mitigating the detrimental waste crisis that leads to climate change.