Convolution operation method and convolution operation device

Abstract

The invention provides a convolution operation method and a convolution operation device. The convolution operation method is used for executing convolution operation on an input feature map to generate a corresponding output feature map, and comprises the steps of dividing each input data block in a plurality of input data blocks into a plurality of non-overlapping areas, and wherein an overlapping area exists between any two adjacent input data blocks; storing a plurality of non-overlapping regions of each input data block into corresponding non-overlapping storage spaces in a cache; generating each input data block according to an area which is stored in the non-overlapped storage space and corresponds to the input data block; and performing convolution operation on the plurality of generated input data blocks to generate the output feature map.

Description

technical field [0001] The present invention relates to a convolution operation method and a convolution operation device, in particular to a convolution operation method and a convolution operation device for dividing an input data block according to the overlap between input data blocks of an input feature map . Background technique [0002] Convolutional Neural Networks (CNN) is currently the main force in the development of deep neural networks, and it is very accurate in image recognition. A typical convolutional neural network includes many layers of operations, such as convolution layers, activation layers, pooling layers, and fully connected layers. [0003] Using a convolution operation module (hardware module, such as a CNN accelerator, etc.) independent of the CPU (Central Processing Unit, central processing unit) can effectively increase the speed of the convolution operation. However, the cache space used to cache the operation data (including input data and c...

AI Technical Summary

Problems solved by technology

However, the cache space used to cache the operation data (including input data and convolution kernel, etc.) in the convolution operation module is limited. When performing the convolution operation, it is impossible to cache all the operation data used by the current convolution layer to the volume In the product operation module

Therefore, if the operation data used in the convolution operation has not been cached in the convolution operation module, the convolution operation module will suspend the convolution operation and load the required operation data from the memory outside the convolution operation module Come in, the convolution operation can only be continued after the required operation data is loaded, which affects the operation speed of the convolution operation module

Images