基于卷積神經(jīng)網(wǎng)絡(luò)的圖像分類模型綜述*
電子技術(shù)應(yīng)用 2023年9月
郭慶梅1,于恒力2,王中訓(xùn)1,劉寧波2
(1.煙臺大學(xué) 物理與電子信息學(xué)院,山東 煙臺 264005;2.海軍航空大學(xué) 信息融合研究所,山東 煙臺 264001)
摘要: 卷積神經(jīng)網(wǎng)絡(luò)在計算機視覺等領(lǐng)域占有一席之地,利用局部連接、權(quán)值共享以及池化操作等特性,有效地提取圖像的局部特征,降低網(wǎng)絡(luò)復(fù)雜度,具有更少的參數(shù)量和更好的魯棒性,因此,吸引了眾多研究者的關(guān)注,使分類模型朝著更輕、更快、更高效的方向迅速發(fā)展。按照卷積神經(jīng)網(wǎng)絡(luò)發(fā)展的時間線,介紹了常用的典型網(wǎng)絡(luò)模型,剖析了其創(chuàng)新點與優(yōu)缺點,并對其未來的發(fā)展方向進行了展望。
中圖分類號:TP183 文獻(xiàn)標(biāo)志碼:A DOI: 10.16157/j.issn.0258-7998.233909
中文引用格式: 郭慶梅,于恒力,王中訓(xùn),等. 基于卷積神經(jīng)網(wǎng)絡(luò)的圖像分類模型綜述[J]. 電子技術(shù)應(yīng)用,2023,49(9):31-38.
英文引用格式: Guo Qingmei,Yu Hengli,Wang Zhongxun,et al. Review of image classification models based on convolutional neural networks[J]. Application of Electronic Technique,2023,49(9):31-38.
中文引用格式: 郭慶梅,于恒力,王中訓(xùn),等. 基于卷積神經(jīng)網(wǎng)絡(luò)的圖像分類模型綜述[J]. 電子技術(shù)應(yīng)用,2023,49(9):31-38.
英文引用格式: Guo Qingmei,Yu Hengli,Wang Zhongxun,et al. Review of image classification models based on convolutional neural networks[J]. Application of Electronic Technique,2023,49(9):31-38.
Review of image classification models based on convolutional neural networks
Guo Qingmei1,Yu Hengli2,Wang Zhongxun1,Liu Ningbo2
(1.School of Physics and Electronic Information, Yantai University, Yantai 264005, China; 2.Information Fusion Institute, Naval Aviation University, Yantai 264001, China)
Abstract: Convolutional neural networks have established themselves as a prominent technique in computer vision and related fields. By leveraging features such as local connections, weight sharing, and pooling operations, these networks are able to effectively extract local features from images, reducing network complexity, and exhibiting fewer parameters and greater robustness. As a result, they have garnered significant attention from researchers and have led to the rapid development of classification models that are lighter, faster, and more efficient. This article presents a timeline of typical network models used in convolutional neural network development, analyzes their innovative points and advantages and disadvantages, and offers insights into their future development directions.
Key words : convolutional neural network;computer vision;feature extraction;classification model
0 引言
卷積神經(jīng)網(wǎng)絡(luò)[1]是一種深度學(xué)習(xí)模型,主要應(yīng)用于圖像和視頻等數(shù)據(jù)的識別與分類。2012年Alex Krizhevsky等人[2]在ImageNet大賽中使用CNN大幅度超越傳統(tǒng)方法,CNN一躍成為計算機視覺領(lǐng)域的熱門技術(shù)。其具有表征學(xué)習(xí)能力、泛化能力以及平移不變性,可以高效處理大規(guī)模圖像且能夠轉(zhuǎn)換成圖像結(jié)構(gòu)的數(shù)據(jù),解決了傳統(tǒng)方法需手動提取特征帶來的耗時、準(zhǔn)確率低等問題,加之計算機性能有了很大的提升[3],使得CNN得到了質(zhì)的發(fā)展,因此在圖像分類、目標(biāo)識別以及醫(yī)療診斷等領(lǐng)域被廣泛應(yīng)用[4],且取得了顯著的成就。
本文詳細(xì)內(nèi)容請下載:http://ihrv.cn/resource/share/2000005634
作者信息:
郭慶梅1,于恒力2,王中訓(xùn)1,劉寧波2
(1.煙臺大學(xué) 物理與電子信息學(xué)院,山東 煙臺 264005;2.海軍航空大學(xué) 信息融合研究所,山東 煙臺 264001)
此內(nèi)容為AET網(wǎng)站原創(chuàng),未經(jīng)授權(quán)禁止轉(zhuǎn)載。