Lightweight method and device for convolutional neural network

Abstract

The embodiment of the invention discloses a lightweight method and device for a convolutional neural network. The method comprises the steps of obtaining a weight coefficient matrix of each convolution kernel in each convolution layer in the convolution neural network model; setting the weight coefficient value of which the absolute value in the weight coefficient matrix is smaller than a preset threshold value to be zero for any one weight coefficient matrix to obtain a first matrix and a second matrix, wherein the first matrix comprises all non-zero weight coefficients in the weight coefficient matrix, and the second matrix comprises the subscript values of all non-zero weight coefficients of the weight coefficient matrix in the weight coefficient matrix; taking the first matrix and thesecond matrix as the compressed weight coefficient matrixes; and carrying out traffic jam identification on the input image according to the weight coefficient matrix after each convolution kernel ineach convolution layer in the convolution neural network model is compressed.

Description

technical field [0001] The present application relates to the technical field of machine learning, in particular to a lightweight method and device for a convolutional neural network. Background technique [0002] In recent years, traffic congestion has become a bottleneck restricting urban economic and social development. It is a concentrated expression of the contradiction between the supply and demand of urban road resources and is closely related to all aspects of urban operation. It directly causes the overall operation efficiency of the city to be low. The short-board effect in the development process is becoming more and more obvious. [0003] The management of traffic congestion should firstly discover congested intersections and road sections in time, and then use new media public information platforms such as traffic broadcasting and microblog to send out congestion information, so that vehicles can choose driving routes reasonably, so as to reduce the pressure on ...

AI Technical Summary

Problems solved by technology

Since the acquisition of various parameters is usually not very accurate, it will lead to inaccurate final processing results, and this method does not have good scalability

As a result, the calculation time is too long, which affects the real-time performance of traffic congestion identification, and thus fails to inform vehicles in a timely manner to choose a reasonable driving route, and has little effect on traffic congestion control.

Images