Understanding deep features with computer-generated imagery

被引:68
作者
Aubry, Mathieu [1 ,2 ]
Russell, Bryan C. [3 ]
机构
[1] Univ Calif Berkeley, Berkeley, CA USA
[2] Univ Paris Est, LIGM, UMR CNRS 8049, ENPC, Paris, France
[3] Adobe Res, Bangalore, Karnataka, India
来源
2015 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV) | 2015年
关键词
D O I
10.1109/ICCV.2015.329
中图分类号
TP18 [人工智能理论];
学科分类号
140502 [人工智能];
摘要
We introduce an approach for analyzing the variation of features generated by convolutional neural networks (CNNs) with respect to scene factors that occur in natural images. Such factors may include object style, 3D viewpoint, color, and scene lighting configuration. Our approach analyzes CNN feature responses corresponding to different scene factors by controlling for them via rendering using a large database of 3D CAD models. The rendered images are presented to a trained CNN and responses for different layers are studied with respect to the input scene factors. We perform a decomposition of the responses based on knowledge of the input scene factors and analyze the resulting components. In particular, we quantify their relative importance in the CNN responses and visualize them using principal component analysis. We show qualitative and quantitative results of our study on three CNNs trained on large image datasets: AlexNet Places and Oxford VGG We observe important differences across the networks and CNN layers for different scene factors and object categories. Finally, we demonstrate that our analysis based on computer generated imagery translates to the network representation of natural images.
引用
收藏
页码:2875 / 2883
页数:9
相关论文
共 43 条
[1]
[Anonymous], 2015, ICLR
[2]
[Anonymous], 2014, P BMVC
[3]
[Anonymous], 2009, TECHNICAL REPORT
[4]
[Anonymous], 2014, ICML
[5]
[Anonymous], 2014, WORKSH INT C LEARN R
[6]
[Anonymous], 2014, CVPR
[7]
[Anonymous], 2015, CVPR
[8]
[Anonymous], ARXIV150303167
[9]
[Anonymous], 2013, ICML
[10]
[Anonymous], ICCV