Convolutional neural network parallel processing method based on OpenCL

Abstract

The invention provides a convolutional neural network parallel processing method based on an OpenCL, and mainly solves the problems of high model complexity and slow operation speed in the existing convolutional neural network parallel processing. The method comprises the steps of obtaining a recombination matrix of an image data matrix; obtaining a weight matrix; carrying out parallel computing on the product of the weight matrix and the recombination matrix in a blocking manner; performing parallel batch normalization on the product matrix; and outputting a characteristic value matrix. According to the invention, a large number of parallel computing units in a computer graphics processing unit (GPU) are utilized to convert the convolution process of the convolutional neural network intolarge-scale matrix multiplication, the product of the weight matrix and the recombination matrix is subjected to block parallel computing, the processing process of convolution layer data is simplified, the access mode of the data is optimized, the reuse rate of the data is improved, and the operation speed of the convolutional neural network is greatly improved.

Description

technical field [0001] The invention belongs to the field of computer technology, and further relates to a convolutional neural network parallel processing method using an open computing language OpenCL (Open Computing Language) in the field of computer vision and deep learning technology. The invention can accelerate the image convolution process of the convolutional neural network, and can be used for real-time target detection of computer vision. Background technique [0002] The convolution process of the convolutional neural network requires a large number of floating-point operations. When the number of layers of the convolutional neural network continues to deepen, the execution efficiency of the CPU is far from meeting the requirements. The GPU provides a large number of parallel computing units. OpenCL can program the GPU on the host side, and the heterogeneous architecture of CPU+GPU can be used to accelerate the image convolution process of the convolutional neura...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to address the above-mentioned deficiencies in the prior art, and propose a method for parallel processing of convolutional neural networks based on OpenCL, which is used to solve the problems of high model complexity and slow running speed in the parallel processing of existing convolutional neural networks

Images