Designing and implementing AI-Enhanced IoT vermicomposting: Revolutionizing waste management with Tiny ML

Abstract

This research proposes an AI-based vermicomposting system that leverages the computational power of a Raspberry Pi to monitor and optimize the growth of Eisenia fetida (red wigglers) for high-quality organic fertilizer production in Rwanda. Traditional vermicomposting methods rely heavily on manual observation or multiple external sensors, which can be expensive and unsuitable for low-resource environments. To address this challenge, the proposed system replaces sensor-based monitoring with periodic image capture and advanced computer vision analysis. A low-cost camera module integrated with the Raspberry Pi captures high-resolution images of the composting bin at regular intervals. These images are analyzed using Convolutional Neural Networks (CNNs) and other image processing techniques to assess worm health, developmental stage, and population density. Based on the visual data, the system provides actionable recommendations for manual interventions to maintain optimal composting conditions. The research adopts a quantitative methodology, employing mathematical models for image classification and statistical methods to validate system performance. Designed for affordability and accessibility, the system utilizes locally available, energy-efficient components to ensure sustainability and ease of adoption. The outcomes of this research are expected to enhance vermicompost quality, promote sustainable agriculture, reduce organic waste, and support climate-resilient farming practices in Rwanda.