Design of Intelligent Classification Compressible Garbage Bin Based on SE-ResNet

To achieve intelligent, efficient, and accurate sorting of household waste, this study proposes an intelligent waste classification device based on an improved ResNet deep neural network integrated with a Squeeze-and-Excitation (SE) attention mechanism, designed to perform waste classification tasks. The dimensions of the sorting mechanism are optimized using the closed-loop vector method, while the servo output power is calculated based on the principle of virtual work, providing a theoretical basis for servo selection. A channel attention mechanism is integrated into the network model to facilitate effective interaction and selection among channels, thereby fully utilizing channel information. The control system employs an Arduino Mega 2560 for sensor signal acquisition and communicates with an NVIDIA Jetson NX via a serial port. Combined with a waste-dropping motion control algorithm, this enables precise control of the motors and servos. Additionally, the system is equipped with a custom-designed compression device that mechanically squeezes compressible waste to improve space utilization. To validate its classification accuracy, twelve common types of household waste were used to test the intelligent classification garbage bins. Experimental results demonstrate that the average waste recognition accuracy of the SE-ResNet-based deep neural network reaches as high as 95.83%, meeting practical application requirements.