TrashScan is an AI-powered surveillance system designed for real-time urban litter detection, density classification, and data visualization. Leveraging CNN-based architectures with Slicing-Aided Hyper Inference (SAHI) and genetic algorithm optimization, it achieves high accuracy in detecting small waste objects across diverse environmental conditions. Integrated with a multi-feed CCTV manager, TrashScan automates alerts for street sweepers and equips local governments with actionable insights, providing a scalable, efficient, and data-driven approach to smarter urban waste management.