Image recognition network construction method and device, equipment and medium

Abstract

The embodiment of the invention provides an image recognition network construction method and device, equipment and a medium, and the method comprises the steps: building an original neural network which comprises multiple layers of network structures, and wherein each layer of network structure comprises a plurality of network modules with search parameters, and a residual branch is arranged between the input end and the output end of each layer of network structure; taking a first image sample set as a training sample, training the original neural network, and updating parameters of the original neural network and search parameters of the plurality of network modules included in each layer of network structure; in the process of training the original neural network, determining a targetnetwork module according to the search parameters of each network module included in each layer of network structure, wherein the target network module is the network module with the maximum current search parameter value in all the network modules; and when a training ending condition is satisfied, determining an image recognition network according to the plurality of target network modules.

Description

technical field [0001] The present invention relates to the technical field of deep learning, in particular to a method, device, equipment and medium for constructing an image recognition network. Background technique [0002] Deep learning technology is widely used in the field of computer vision due to its advantages of fast speed and high accuracy. For deep learning, the design of neural network architecture is the most important factor in determining its effect. Due to the imperfect development of current deep learning theory, designing an ideal neural network architecture requires not only rich knowledge and in-depth understanding of the designer, but also a large number of exploratory experiments. This requires a lot of time cost, human cost and computing resources. In recent years, NAS (Neural Architecture Search, Neural Network Architecture Search), as a method for automatically designing network architectures, has been applied to better solve the above problems. ...

AI Technical Summary

Problems solved by technology

Although the search speed of this method is fast and the effect is good, it can only search shallow networks, and this method takes up a lot of video memory, and cannot search for larger network structures, resulting in the searched image recognition network is not the best. network structure

Images