An Intelligent Garbage Classification System Using a Lightweight Network MobileNetV2

In response to the current problems of low garbage disposal efficiency and high labor cost, this paper proposes an intelligent and low-cost garbage classification system. By making a dataset, the system is trained with the MobileNetV2 network for migration learning under the Keras framework to recognize and classify garbage types, and deployed in embedded equipment to achieve accurate classification of individual garbage. The system has functions such as full load detection, voice broadcast, and recognition of garbage count. When the system detects the light intensity below a certain threshold, the lighting system above the system automatically turns on to make the system suitable for the night environment. The experimental results show that the system can operate normally, with a prediction time of 62ms for garbage recognition, using 292.9K RAM and 212.9K Flash, and an accuracy rate of 98.7% for garbage recognition.