Pixels to Policy: A Multi-Modal AI Framework for Proactive, Sustainable Waste Management Across Urban & Rural Regions
Year: 2025
Type: dissertation
Abstract: The rapid growth of global waste, driven by urbanization and population expansion, contributes to the increasing demands of smarter, faster, and more adaptable waste management solutions. This research proposes an AI-powered, multi-modal framework that unifies waste detection using YOLOv8s, trend/pattern forecasting using LSTM, and suggestive policy generation using Zephyr-7b-beta into a single, automated system. By analyzing images of waste, predicting future generation patterns for specific locations, and producing tailored, actionable strategies, the system empowers municipalities and communities to take proactive measures before waste issues escalate. This solution is designed to operate in both urban and rural environments, including regions where the data is scarce while shifting from reactive waste management responses to anticipatory interventions. The suggestive policies that are generated are context-specific, addressing regional needs, available resources, and sustainability goals, and are evaluated for relevance, feasibility, and innovation. Real-world validation of the solution demonstrates the framework’s ability to bridge the gap between raw environmental data and practical decision-making, offering a scalable, location-agnostic solution that supports Sustainable Development Goals and enables more efficient, responsible, and inclusive waste management practices.
Author: Singh, Devanshu
Language: English
Subfield: Waste Management and Disposal
Field: Environmental Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7204611289