An FPGA Implementation Method of Lightweight Deep Convolutional Neural Network

Abstract

The present invention proposes an FPGA implementation method of a lightweight deep convolutional neural network, the steps of which are: constructing a lightweight deep convolutional neural network; initializing a lightweight deep convolutional neural network; generating a training set and a verification set; Train lightweight deep convolutional neural networks; design the basic components of lightweight convolutional neural networks in FPGAs; implement trained lightweight deep convolutions in FPGAs product neural network. The present invention simplifies the network structure of the deep convolutional neural network by building a lightweight deep convolutional neural network, optimizes the training method of the lightweight deep convolutional neural network, and improves the implementation in the Field Programmable Logic Gate Array FPGA. Resource utilization of deep convolutional neural networks.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to an FPGA implementation method of a lightweight deep convolutional neural network in the technical field of digital image pattern recognition. The invention can be used to realize a lightweight deep convolutional neural network on a field programmable logic gate array FPGA (Field Programmable Gate Array). Background technique [0002] In recent years, with the rapid development of deep learning, artificial intelligence has gradually been integrated into various fields. Among them, the algorithm model based on deep convolutional neural network DCNN (Deep Convolutional Neural Network) has the advantage of higher accuracy compared with other deep learning network frameworks, and has been widely concerned since it was proposed in 2015. So far, the deep convolutional neural network model is one of the most abundant neural network models, but it is rarely implemented on ...

AI Technical Summary

Problems solved by technology

[0004] The object of the present invention is to aim at the deficiencies in the above-mentioned prior art, propose a kind of FPGA implementation method of lightweight deep convolutional neural network, be used to solve the method for realizing deep convolutional neural network in programmable logic gate array FPGA When the level of the network structure is deep, the resource occupation of field programmable logic gate array FPGA is too high for technical problems

Images