DFT-based transformation invariant pooling layer for visual classification
- Authors
- Ryu, Jongbin; Yang, Ming-Hsuan; Lim, Jongwoo
- Issue Date
- Oct-2018
- Publisher
- Springer Verlag
- Citation
- Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), v.11218 LNCS, pp.89 - 104
- Indexed
- SCOPUS
- Journal Title
- Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
- Volume
- 11218 LNCS
- Start Page
- 89
- End Page
- 104
- URI
- https://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/149307
- DOI
- 10.1007/978-3-030-01264-9_6
- ISSN
- 0302-9743
- Abstract
- We propose a novel discrete Fourier transform-based pooling layer for convolutional neural networks. The DFT magnitude pooling replaces the traditional max/average pooling layer between the convolution and fully-connected layers to retain translation invariance and shape preserving (aware of shape difference) properties based on the shift theorem of the Fourier transform. Thanks to the ability to handle image misalignment while keeping important structural information in the pooling stage, the DFT magnitude pooling improves the classification accuracy significantly. In addition, we propose the DFT+ method for ensemble networks using the middle convolution layer outputs. The proposed methods are extensively evaluated on various classification tasks using the ImageNet, CUB 2010-2011, MIT Indoors, Caltech 101, FMD and DTD datasets. The AlexNet, VGG-VD 16, Inception-v3, and ResNet are used as the base networks, upon which DFT and DFT+ methods are implemented. Experimental results show that the proposed methods improve the classification performance in all networks and datasets.
- Files in This Item
-
Go to Link
- Appears in
Collections - 서울 공과대학 > 서울 컴퓨터소프트웨어학부 > 1. Journal Articles
![qrcode](https://api.qrserver.com/v1/create-qr-code/?size=55x55&data=https://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/149307)
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.