{"title":"An improved compression algorithm based on IDN model of image super-resolution reconstruction","authors":"Zemin Xu, Jian Xu, Bing Song, Zhengguang Xie","doi":"10.1117/12.2639111","DOIUrl":null,"url":null,"abstract":"At present, the convolutional neural network is deepening in level and has a huge amount of computation, so it is difficult to realize application in scenarios with low computing capacity. Therefore, this paper proposes a method based on channel pruning and weight quantization to reduce the amount of computation and compress the image super-resolution to reconstruct the network model IDN. Experimental results show that the proposed method effectively compresses the model structure, greatly shortens the calculation time of the model and makes the model more lightweight under the premise that the performance indexes are basically unchanged.","PeriodicalId":336892,"journal":{"name":"Neural Networks, Information and Communication Engineering","volume":"102 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Neural Networks, Information and Communication Engineering","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1117/12.2639111","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
At present, the convolutional neural network is deepening in level and has a huge amount of computation, so it is difficult to realize application in scenarios with low computing capacity. Therefore, this paper proposes a method based on channel pruning and weight quantization to reduce the amount of computation and compress the image super-resolution to reconstruct the network model IDN. Experimental results show that the proposed method effectively compresses the model structure, greatly shortens the calculation time of the model and makes the model more lightweight under the premise that the performance indexes are basically unchanged.