Deep Transfer Learning-Based Waste Classification Research

Abstract

To address the problems of poor manual detection environment, error-prone, difficulty and low efficiency of garbage classification, a method of domestic garbage classification using deep transfer learning is proposed. Firstly, image datasets for garbage classification are constructed while data augmentation, secondly, deep convolutional neural networks ResNeXt and MobileNetV2 are built to fine-tune the network transfer parameters to suit the garbage classification task, and finally, the effects of network freezing layers and learning rate on the network structure caused by different magnitudes are explored under the convolutional neural networks based on deep migra-封皓元