Neural network acceleration system based on FPGA

Abstract

The invention relates to a neural network acceleration system based on FPGA. According to the neural network acceleration system, on the basis of the natural parallelism of a convolutional neural network and the sparsity of a full connection layer, through operation resource reuse, parallel data processing and assembly line design, a sparse matrix multiplier is designed by utilizing the sparsity of the full connection layer, so the operation speed is greatly improved; the use of resources is reduced; and the inference speed is improved under the condition that the inference accuracy of the convolutional neural network is not influenced. According to the invention, through operation resource reuse, parallel data processing and assembly line design, the sparse matrix multiplier is designed by utilizing the sparsity of the full connection layer, so the operation speed is greatly improved; the use of resources is reduced; and under the condition that the inference accuracy of the convolutional neural network is not influenced, the overall power consumption of the system is reduced, and the inference speed is improved.

Description

technical field [0001] The invention belongs to the fields of artificial intelligence and electronics, and in particular relates to an FPGA-based neural network acceleration system. Background technique [0002] With the rapid development and wide application of deep learning in recent years, convolutional neural network (CNN) has become the best method in the field of detection and recognition. It can automatically learn to extract features from data sets, and the more network layers , the extracted features are more global. Through local connection and weight sharing, the generalization ability of the model can be improved, and the accuracy of recognition and classification can be greatly improved. [0003] And with the development of the Internet of Things, deploying the convolutional neural network on the embedded side has to process a large amount of data, which will consume a lot of resources and energy, and embedded devices usually use batteries to maintain work, and...

AI Technical Summary

Problems solved by technology

[0003] And with the development of the Internet of Things, deploying the convolutional neural network on the embedded side has to process a large amount of data, which will consume a lot of resources and energy, and embedded devices usually use batteries to maintain work, and frequent battery replacement will increase costs, so It has important practical significance for the acceleration of calculation in the inference stage and low power consumption design

Images