The Research On Image Recognition Algorithm Based On Convolutional Neural Networks

Convolutional Neural Networks is an identification algorithm,which is based on the artificial neural network and combined with deep learning.So far it has been successfully applied to object detection,image recognition,speech recognition and other fields.Although the identification technology of Convolutional Neural Networks has gradually stabilized,it is mainly implemented in single machine serial mode and there are many problems such as too long training time and lack of memory.In this paper,through the analysis of the Convolution Neural Networks algorithm and the contrast of the image recognition technology,the Convolution Neural Networks algorithm is improved to enhance the effect of image recognition.The primary coverage includes multi region image recognition and the design of Convolutional Neural Networks algorithm based on MapReduce.The multi region image recognition method is based on the traditional Convolution Neural Networks.When the network structure is simplified and the image recognition is tested,a plurality of test regions are obtained according to the regional division,and the accuracy of image recognition is improved;On this basis,the Convolutional Neural Networks algorithm is designed based on MapReduce platform,the Map function is used to divide the data into N modules and is processed to form the intermediate data,and then it is collected by the Reduce function,the training speed is faster through the parallel computing of MapReduce.Finally,the improved Convolutional Neural Networks algorithm is programmed in the MapReduce environment,and by comparing with the experiment results of traditional Convolution Neural Networks,it is concluded that the improved algorithm can improve the accuracy and training speed of image recognition.