Intelligent Waste Classification System Design and Implementation Based on Deep Learning

With the acceleration of urbanization and the improvement of environmental protection awareness, garbage classification and waste disposal have become the focus of social attention. This paper aims to solve the problem of garbage classification in the process of social production and life, including reducing the pressure of front-end garbage collection and increasing garbage disposal efficiency. This paper uses mobilenetv4 for feature extraction, transfer learning based on the pre-trained model of ImageNet12k dataset, and classifies garbage into 4 categories and 35 subcategories. Combined with embedded technology, the trained model is deployed on Raspberry Pi 4B to control the hardware to realize the entire garbage classification process. After testing, the improved recognition accuracy can reach 0.93, and an average of 0.23s can be completed once, which provides an effective solution for realizing image recognition on low-cost mobile devices with limited computing power, and can reduce the workload of manual classification, improve classification efficiency, and contribute to the development of garbage classification.