Constructing Fast Network through Deconstruction of Convolution

Yunho Jeon, Junmo Kim
5/28/2018
cs.CV

Official code: jyh2986/Active-Shift (33 stars, last commit 2019-03-20).

Abstract

Convolutional neural networks have achieved great success in various vision tasks; however, they incur heavy resource costs. By using deeper and wider networks, network accuracy can be improved rapidly. However, in an environment with limited resources (e.g., mobile applications), heavy networks may not be usable. This study shows that naive convolution can be deconstructed into a shift operation and pointwise convolution. To cope with various convolutions, we propose a new shift operation called active shift layer (ASL) that formulates the amount of shift as a learnable function with shift parameters. This new layer can be optimized end-to-end through backpropagation and it can provide optimal shift values. Finally, we apply this layer to a light and fast network that surpasses existing state-of-the-art networks.

View on arXivView PDF

Code Implementations(1)

jyh2986/Active-ShiftOfficial100%

Caffe implementation for Active Shift Layer(ASL)

334C++, CudaSep 10, 20187 years agoNOASSERTION
aslcaffeconvolutional-layersdeeplearning

Similar Papers

All You Need is a Few Shifts: Designing Efficient Convolutional Neural Networks for Image Classification

87% match1 repo152019

SqueezeBERT: What can computer vision teach NLP about efficient neural networks?

82% match1 repo162.1k2020

Learning Filter Basis for Convolutional Neural Network Compression

81% match1 repo182019

MobileViT: Light-weight, General-purpose, and Mobile-friendly Vision Transformer

81% match1 repo2.0k2021

Run, Don't Walk: Chasing Higher FLOPS for Faster Neural Networks

81% match2 repos8352023

GhostShiftAddNet: More Features from Energy-Efficient Operations

81% match1 repo62021

Cite this paper

@article{jeon2018constructing,
  title  = {Constructing Fast Network through Deconstruction of Convolution},
  author = {Yunho Jeon and Junmo Kim},
  year   = {2018},
  eprint = {1806.07370},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1806.07370}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1806.07370.svg)](https://codeofpaper.com/paper/1806.07370)

Discussion