Open AccessDesign and Implementation of Occlusion Image Recognition Algorithm Based on Deep Convolution Generative Adversarial Network
Author(s) -
Zhijun Liu,
Chao Feng,
Xuefeng Pan
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1883/1/012045
Subject(s) - artificial intelligence , convolution (computer science) , computer science , image (mathematics) , sample (material) , feature (linguistics) , pattern recognition (psychology) , feature extraction , algorithm , handwriting , process (computing) , computer vision , artificial neural network , linguistics , chemistry , philosophy , chromatography , operating system
In the image recognition process, the accuracy of image recognition is affected due to partial coverage or light problems. This paper proposes a method to solve the problem of target occlusion based on a deep convolution generation confrontation network. The method actively occludes the feature map after the feature is extracted to generate the confrontation sample, and generates a map and a mask from the input image. At the same time, a method is proposed. This new loss function applies the algorithm to handwriting recognition and trains the network through large-scale sample data sets. Experiments show that this method significantly improves the accuracy of image extraction.