Research On Classification Method Of Domestic Waste Image Based On Deep Learning

How to use a reasonable way to classify domestic waste is one of the important problems that need to be solved urgently in contemporary society.Using deep learning technology to classify domestic garbage images has great potential.At present,there are not many researches on the classification of domestic waste images using deep learning technology.The only research work shows that the performance of domestic waste detection models needs to be improved,and the types and quantities of waste in the public domestic waste data set used are small,leading to deep learning model having problems such as overfitting or insufficient training.In order to solve the above problems,this paper firstly studies the DCGAN-based enhancement algorithm for the domestic waste data set,and then proposes a new target detection model for the domestic waste target detection task.The specific work of this paper includes:(1)Propose a DCGAN-based enhancement method for household garbage data setAiming at the task of enhancing household garbage data,this paper takes garbage＿classify data set as the research object,and studies the method of enhancing household garbage data based on DCGAN.First,the DCGAN network structure is finetuned,and then the original training set is input into the network,and the generated image is integrated with the original training set to obtain the expanded training set.Secondly,this article uses the expanded training set to train and learn classification tasks in the three neural network models of Alex Net,VGG and Res Net.The experimental results show that compared with the traditional data enhancement method based on geometric transformation image operation,the average accuracy of classification is increased by4.09%.Compared with the method without data enhancement,the average accuracy of classification is increased by 7.27%.It is verified that this method can effectively expand the data set,and it can be used as a technology for enhancing data of domestic garbage.(2)Propose an improved YOLOv3 domestic garbage detection model——YOLOTrashAiming at the task of detecting domestic garbage targets,based on the previously expanded data as the research object,this paper proposes an improved YOLO-Trash domestic garbage image detection model based on YOLOv3.Based on the YOLOv3 model,the YOLO-Trash model introduces the Dense block module,Mish activation function and CBAM attention mechanism.The introduction of the above modules is to allow the model to learn deeper image features and to strengthen the model’s non-linear expression ability.The experimental results show that compared with YOLOv3,the accuracy of the proposed model is increased by 2.4%,the recall rate is increased by 8.0%,the F1 score is increased by 4.2%,and the m AP0.5 score is increased by 5.1%.It is verified that this method improves the detection performance of the detection model,and can better detect and classify domestic garbage in more complex life scenes.