Energy-Efficient and Comprehensive Garbage Bin Overflow Detection Model Based on Spiking Neural Networks

With urbanization and population growth, waste management has become a pressing issue. Intelligent detection systems using deep learning algorithms to monitor garbage bin overflow in real time have emerged as a key solution. However, these systems often face challenges such as lack of dataset diversity and high energy consumption due to the extensive use of IoT devices. To address these challenges, we developed the Garbage Bin Status (GBS) dataset, which includes 16,771 images. Among them, 8408 images were generated using the Stable Diffusion model, depicting garbage bins under diverse weather and lighting scenarios. This enriched dataset enhances the generalization of garbage bin overflow detection models across various environmental conditions. We also created an energy-efficient model called HERD-YOLO based on Spiking Neural Networks. HERD-YOLO reduces energy consumption by 89.2% compared to artificial neural networks and outperforms the state-of-the-art EMS-YOLO in both energy efficiency and detection performance. This makes HERD-YOLO a promising solution for sustainable and efficient urban waste management, contributing to a better urban environment.