Application Of Convolutional Neural Network In Image Classification Under Tensorflow Framework

With the rapid development of AI,AI technology keeps in touch with people's life and has made outstanding contributions in different fields.Artificial intelligence has become an important research topic nowadays,not only in algorithm and hardware technology,but also in practical application.Deep learning is an important part of artificial intelligence.It has become the main solution in image recognition,language recognition and natural language processing.On the basis of traditional algorithm,deep convolution neural network is invented to meet the feature extraction and learning requirements of a large number of data.Distributed computing and cloud computing are proposed in hardware configuration to solve the difficult problem that deep learning requires high training environment configuration,so that more people can join in deep learning.In the study of deep learning.Many of the world's top hightech companies set up laboratories to find a more convenient and rapid development model for in-depth learning,resulting in the Tensorflow in-depth learning framework studied in this paper.Cloud computing like Alibaba Cloud Computing also provides a cloud platform that can meet the needs of in-depth learning.In this paper,the deep convolution neural network model is studied,applied and improved by using cloud computing capability in hardware environment and Tensorflow deep learning framework based on Python language.Firstly,the background,research significance and research status at home and abroad are introduced.Then the basic theory of convolution neural network is elaborated,and three important convolution neural network models are introduced.Afterwards,we introduce the Alibaba Cloud Computing platform's configuration of computing environment for in-depth learning and the principle and environment of Tensorflow's in-depth learning framework.Finally,on the basis of the former principle and environment,different convolutional neural network models are used to simulate CIAR-10 data set and driver driving condition detection data set.According to the output of the experimental results,the reasons for the influence of the results and the improvement methods are analyzed theoretically,and the network is further trained through the improved method.To improve the data effect.Complete the ultimate goal of data set classification.