Regularization method and system based on class activation mapping graph guidance

Abstract

The invention discloses a regularization method and a regularization system based on class activation mapping graph guidance. The method comprises the following steps: S1, generating a channel weightfactor and a class activation mapping graph based on a label class by using global pooling layer and full connection layer parameters of a deep neural network; S2, sorting the generated channel weightfactors and the class activation mapping graphs according to contribution degrees of the channel weight factors and the class activation mapping graphs to all channels and space regions of each layerof feature graphs in the neural network; s3, obtaining a feature channel set and a feature point set according to the step S2, and further obtaining binary mask graphs M(1) and M(2) based on a channel weight factor and a class activation mapping graph; s4, generating a random seed binary image M(3) based on Bernoulli distribution, performing logical operation with M(1) and M(2) to obtain a finalbinary mask image M, and thus obtaining a regularized mask image M1; and S5, iteratively carrying out training processes from S1 to S4 for multiple times to complete a regularization optimization process.

Description

technical field [0001] The invention relates to the field of computer vision based on deep learning, in particular to a regularization method and system guided by a class activation map. Background technique [0002] In recent years, with the increasing development of massive data and deep learning, various visual recognition tasks have made great progress. However, although deep neural networks bring superior performance, due to the lack of interpretability, the prediction results of deep neural networks are often unconvincing, and also leave potential security risks. Taking automatic driving as an example, an automatic driving system with poor generalization ability may cause major traffic safety accidents and cause personal and property losses. Therefore, how to make the neural network have a strong representation ability and build a more robust system becomes very important. [0003] The regularization method is the most commonly used method in the field of deep learni...

AI Technical Summary

Problems solved by technology

However, although these methods do start from the image itself and consider the strong practical meaning of the image itself, these methods do not achieve robustness and interpretability in the true sense.

Images