Scene understanding based on Multi-Scale Pooling of deep learning features

DOI

10.2991/amcce-15.2015.308How to use a DOI?

Keywords

CNNs; MOP-CNN; SPP-net; Scenes understanding

Abstract

Deep convolutional neural networks (CNNs) have recently shown impressive performance as generic representation for recognition. However, the feature extracted f¬rom global CNNs lack geometric invariance, which limits their robustness for classification and detection of highly variable objects .To improve the invariance of the features w¬ithout degrading their discriminative power and speed up the calculation, we follow t¬he next two method. Firstly, we adopt the scheme called multi-scale orderless pooling (MOP-CNN) which extracts CNNs activation from local patches of the image at multiple scale levels, performs orderless VLAD pooling of these activations at each level separately, and concatenates the result. Second, to speed up the calculation, we adapt the SPP-net as the CNNs architecture. Using SPP-net, we compute the feature maps from the entire image only once, and then pool features in arbitrary regions (sub-images) to generate fi¬xed-length representations for training the detectors. This method avoids repeatedly computing the convolu-tional features. On the challenging SUN397 Scenes classification datasets, our method achieves competitive classification results.

Copyright

© 2015, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - DongYang Li
AU  - Yue Zhou
PY  - 2015/04
DA  - 2015/04
TI  - Scene understanding based on Multi-Scale Pooling of deep learning features
BT  - Proceedings of the 2015 International Conference on Automation, Mechanical Control and Computational Engineering
PB  - Atlantis Press
SN  - 1951-6851
UR  - https://doi.org/10.2991/amcce-15.2015.308
DO  - 10.2991/amcce-15.2015.308
ID  - Li2015/04
ER  -