Convolution operation device and convolution operation method

Abstract

The subject of the invention is a convolution operation device and a convolution operation method. The convolution operation device includes a convolution operation module, a memory and a buffer device. The convolution operation module has multiple convolution units. Each convolution unit carries out a convolution operation on the basis of a filter and multiple pieces of current data, and retainsa part of the current data after the convolution operation. The buffer device is in coupling connection with the memory and the convolution operation module, obtains multiple pieces of new data from the memory, and inputs the new data to the convolution units, and the new data do not overlap with the current data.

Description

technical field [0001] The present invention relates to a convolution operation device, in particular to a convolution operation device suitable for data streams. Background technique [0002] Deep learning is already one of the important application technologies for artificial intelligence (AI). Among them, Convolutional Neural Network (CNN) is a deep learning efficient recognition technology that has attracted widespread attention in recent years. Compared with other deep learning architectures, especially in the field of pattern classification, such as image and speech recognition, the convolutional neural network avoids complicated pre-processing, and can directly input the original image or raw data, thus obtaining more It is widely used and can achieve better recognition results. [0003] However, the convolution operation is a very performance-consuming operation. In the application of the convolutional neural network, especially the convolution operation and contin...

AI Technical Summary

Problems solved by technology

[0003] However, convolution operation is a very performance-intensive operation. In the application of convolutional neural n...

Images