A target classification method based on a convolutional neural network

Abstract

The invention provides a target classification method based on a convolutional neural network. the target image is directly traversed without adopting a convolution kernel; instead, a sliding window with the same size as the output feature image is adopted to traverse the whole target image row by row and column by column; pixel points corresponding to the target image are extracted to serve as sub-images; correspondingly each characteristic parameter of the convolution kernel is multiplied by each sub-image to obtain an intermediate image; and finally, the sum value of the intermediate imagesis taken as an output feature image. Under the premise that a convolution result which is the same as that of an existing convolution implementation mode is obtained, convolution operation is dividedinto multiplication and addition operation of a single point, the number of times of address hopping when a microprocessor reads data in the convolution implementation process can be reduced to the maximum extent, and then the hardware processing efficiency is greatly improved.

Description

technical field [0001] The invention belongs to the technical field of image classification, in particular to an object classification method based on a convolutional neural network. Background technique [0002] Image classification is an important category in the field of image processing technology. How to quickly and accurately realize image classification is a very popular research topic in the current image field. In the past five years, convolutional neural networks have achieved good results in the fields of image feature extraction, classification and recognition. [0003] Since the convolutional neural network needs to sample the convolution kernel to traverse the input feature map when performing convolution calculations, every time a traversal is performed, an address jump will occur. Among them, the address jump refers to the process of convolution calculation. The address jump operation is performed on the input feature map data stored sequentially, and it tak...

AI Technical Summary

Problems solved by technology

[0003] Since the convolutional neural network needs to sample the convolution kernel to traverse the input feature map when performing convolution calculations, every time a traversal is performed, an address jump will occur. Among them, the address jump refers to the process of convolution calculation. The address jump operation is performed on the input feature map data stored sequentially, and it takes more logic control to implement the address jump operation on the hardware platform than on the software platform. As the number of traversals increases, ARM (microprocessor) NEON (vector Co-processing unit) when reading data, the address jump will also increase exponentially, which greatly reduces the efficiency of hardware processing

Images