Real-time Large data image classification under MapReduce frame

Real-time Large data image classification under MapReduce frame

Zhang Feng Lin Wang Sheng Blue

Image data as an important part of large data contains rich knowledge, and image classification has a wide range of applications, the use of traditional classification methods can not meet the needs of real-time computing. In order to solve this problem, a parallel online extreme learning machine algorithm is proposed. Firstly, the output weights matrix of the hidden layer is obtained by using the Secondly, the matrix is segmented according to the characteristics of the MapReduce computing frame. In lieu of the original large-scale matrix accrual operation, the segmented matrices are computed in parallel on different working nodes, and the final classifier is obtained by merging the result key values on the compute nodes. Under the precondition of guaranteeing the original calculation precision, In this paper, the algorithm is extended in the MapReduce frame, and the results of classification of large scale image data with face image as example show that the algorithm can classify large data images quickly and accurately.

Real-time Large data image classification under MapReduce frame