| 112 | 0 | 143 |
| 下载次数 | 被引频次 | 阅读次数 |
卷积神经网络(Convolutional Neural Network,CNN)是一种常用的图像特征提取方法,它以更为紧凑的表示将图像转换为特征空间,其更具鉴别性和鲁棒性的特征表述能力为图像分类识别提供了一种新颖而有前景的解决方案。然而,现有的CNN层内骨干网络缺乏对于每个分组通道内部的信息的融合,不能充分挖掘各分组通道内部特征的联系。提出新的层内优化网络Skip2Net,在Res2Net的基础上引入了一个跳跃连接,有效地聚合了通道内的高级语义特征和低级空间细节,实现了特征重用。
Abstract：Convolutional Neural Network(CNN) is a commonly used image feature extraction method, which converts an image into a feature space with a more compact representation. Its more discriminative and robust feature expression ability provides a novel and promising solution for image classification and recognition. However, the existing backbone network in the CNN layer lacks the fusion of the information within each grouping channel, and cannot fully explore the relationship between the internal characteristics of each grouping channel. In order to solve this problem, this paper proposes a new intra-layer optimization network Skip2 Net, which introduces a skip connection on the basis of Res2 Net, thus effectively aggregating the high-level semantic features and low-level spatial details in the channel, and achieving the feature reuse.
[1]He K,Zhang X,Ren S,et al.Deep residual learning for image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2016:770-778.
[2]Huang G,Liu Z,Van Der Maaten L,et al.Densely connected convolutional networks[C]//Proceedings of the IEEEconference on computer vision and pattern recognition.2017:4700-4708.
[3]Ronneberger O,Fischer P,Brox T.U-net:Convolutional networks for biomedical image segmentation[C]//International Conference on Medical image computing and computer-assisted intervention.Springer,Cham,2015:234-241.
[4]Krizhevsky A,Sutskever I,Hinton G E.Imagenet classification with deep convolutional neural networks[C]//Advances in neural information processing systems.2012:1097-1105.
[5]Szegedy C,Liu W,Jia Y,et al.Going deeper with convolutions[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2015:1-9.
[6]Chollet F.Xception:Deep learning with depthwise separable convolutions[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2017:1251-1258.
[7]Szegedy C,Ioffe S,Vanhoucke V,et al.Inception-v4,inception-resnet and the impact of residual connections on learning[C]//Thirty-first AAAI conference on artificial intelligence.2017:597-605.
[8]Gao S,Cheng M M,Zhao K,et al.Res2net:A new multiscale backbone architecture[J].IEEE transactions on pattern analysis and machine intelligence,2019:135-142.
[9]Xie S,Girshick R,Dollár P,et al.Aggregated residual transformations for deep neural networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2017:1492-1500.
[10]Y.Le Cun et al.Backpropagation applied to handwritten zip code recognition[J]Neural Computer,1989(4):541-551.
[11]Zhang X,Zhou X,Lin M,et al.Shufflenet:An extremely efficient convolutional neural network for mobile devices[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2018:6848-6856.
[12]宋晓茹,吴雪,高嵩,陈超波.基于深度神经网络的手写数字识别模拟研究[J].科学技术与工程.2019(5):193-197.
[13]陈超,齐峰.卷积神经网络的发展及其在计算机视觉领域中的应用综述[J].计算机科学.2019(3):63-72.
基本信息:
中图分类号:TP391.41;TP183
引用信息:
[1]黄颖聪,孟凡阳.基于CNN层内结构优化的图像分类[J].深圳信息职业技术学院学报,2020,18(06):37-43.
基金信息:
国家自然科学基金（项目编号：61871154）;; 深圳市科技计划项目（项目编号：KJYY20170724152625446）
2020-12-15
2020-12-15