计算机科学 ›› 2017, Vol. 44 ›› Issue (Z6): 146-150.doi: 10.11896/j.issn.1002-137X.2017.6A.034
俞海宝,沈琦,冯国灿
YU Hai-bao, SHEN Qi and FENG Guo-can
摘要: 经典的反卷积可视化模型通过反池化、反激活、反卷积将特征图像还原至原图像空间,可视化网络节点从输入图像学习到的特征,有助于探究卷积神经网络运行良好的机制,但是由于采用近似处理,还原特征不明显。本研究引入数值求解方法来代替原模型中直接用卷积核的反转近似反卷积核的方法。先构造数据集:随机生成大小、形状、位置不一的结构简单、角点特征明显的三角形和矩形,用于组成层次结构逐渐复杂的数据集,并利用该数据集测试模型的可视化效果。实验表明,改进后的可视化模型提取的特征更明显,引入的噪音更少,可以更为精确地将激活网络节点从原图像学习的特征可视化。在更大的数据库上进行实验来验证结果,并利用这种结果进一步探究准确率与网络结构之间存在何种关系。
[1] LECUN Y,BOTTOU L,BENGIO Y,et al.Gradient-basedlearning applied to document recognition[J].Proceedings of the IEEE,1998,86(11):2278-2324. [2] LECUN Y,BENGIO Y,HINTON G.Deep learning[J].Nature,2015,521(7553):436-444. [3] KRIZHEVSKY A,SUTSKEVER I,HINTON G E.Imagenetclassification with deep convolutional neural networks[C]∥Advances in Neural Information Processing Systems.2012:1097-1105. [4] 王晓刚.深度学习在图像识别中的研究进展与展望.http://chuangsong.me/n/2103247. [5] OUYANG W L,WANGX G.Joint deep learning for pedestrian detection[C]∥ICCV.New York:IEEE,2013:2056-2063. [6] CHEN Y N,HAN C C,WANG C T,et al.The application of a convolution neural network on face and license plate detection[C]∥The 18th International Conference on Pattern Recognition.Hongkong:IEEE,2006,4:552-555. [7] NGUYEN A,YOSINSKI J,CLUNE J.Deep neural networksare easily fooled:High confidence predictions for unrecognizable images[C]∥2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).IEEE,2015:427-436. [8] FAWZI A,FAWZI O,FROSSARD P.Analysis of classifiers’robustness to adversarial perturbations[J/OL].Available: http://arxiv.org/abs/1502.02590. [9] SCHWENK H,BOUGARER F,BARRAULT L.Efficient trai-ning strategies for deep neural network language models[C]∥NIPS workshop on Deep Learning and Representation Learning.2014. [10] ZEILER M D,KRISHNAN D,TAYLOR G W,et al.Deconvolutional networks[C]∥2010 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).IEEE,2010:2528-2535. [11] KAVUKCUOGLU K,SERMANET P,BOUREAU Y,et al.Learning onvolutional feature hierarchies for visual recognition[C]∥Proceedings of Advances in Neural Information Proces-sing Systems,2010.New York:Curran Associates,Inc.,2010:1090-1098. [12] ZEILER M D,TAYLOR G W,FERGUS R.Adaptive deconvolutional networks for mid and high level feature learning[C]∥2011 IEEE International Conference on Computer Vision (ICCV).IEEE,2011:2018-2025. [13] DOSOVITSKIY A,BROX T.Inverting Visual Representations with Convolutional Networks[J].arXiv preprint arXiv:1506.02753,2015. [14] ZEILER M D,FERGUS R.Visualizing and understanding con-volutional networks[M]∥Computer Vision-ECCV 2014.Sprin-ger International Publishing,2014:818-833. [15] YOSINSKI J,CLUNE J,NGUYEN A,et al.Understandingneural networks through deep visualization[J].arXiv preprint arXiv:1506.06579,2015. [16] BOUREAU Y L,PONCE J,LECUN Y.A Theoretical Analysis of Feature Pooling in Visual Recognition[C]∥International Conference on Machine Learning.2010:459-459. [17] FARABET C,COUPRIE C,NAJMAN L,et al.Learning hierarchical features for scene labeling[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2013,35(8):1915-1929. |
No related articles found! |
|