AsymmNet: Towards ultralight convolution neural networks using asymmetrical bottlenecks

04/15/2021
by   Haojin Yang, et al.
0

Deep convolutional neural networks (CNN) have achieved astonishing results in a large variety of applications. However, using these models on mobile or embedded devices is difficult due to the limited memory and computation resources. Recently, the inverted residual block becomes the dominating solution for the architecture design of compact CNNs. In this work, we comprehensively investigated the existing design concepts, rethink the functional characteristics of two pointwise convolutions in the inverted residuals. We propose a novel design, called asymmetrical bottlenecks. Precisely, we adjust the first pointwise convolution dimension, enrich the information flow by feature reuse, and migrate saved computations to the second pointwise convolution. By doing so we can further improve the accuracy without increasing the computation overhead. The asymmetrical bottlenecks can be adopted as a drop-in replacement for the existing CNN blocks. We can thus create AsymmNet by easily stack those blocks according to proper depth and width conditions. Extensive experiments demonstrate that our proposed block design is more beneficial than the original inverted residual bottlenecks for mobile networks, especially useful for those ultralight CNNs within the regime of <220M MAdds. Code is available at https://github.com/Spark001/AsymmNet

READ FULL TEXT

page 1

page 2

page 3

page 4

research
09/28/2019

Training convolutional neural networks with cheap convolutions and online distillation

The large memory and computation consumption in convolutional neural net...
research
07/05/2020

Rethinking Bottleneck Structure for Efficient Mobile Network Design

The inverted residual block is dominating architecture design for mobile...
research
06/10/2019

BlockSwap: Fisher-guided Block Substitution for Network Compression

The desire to run neural networks on low-capacity edge devices has led t...
research
06/13/2017

SEP-Nets: Small and Effective Pattern Networks

While going deeper has been witnessed to improve the performance of conv...
research
12/21/2022

Revisiting Residual Networks for Adversarial Robustness: An Architectural Perspective

Efforts to improve the adversarial robustness of convolutional neural ne...
research
09/05/2022

LKD-Net: Large Kernel Convolution Network for Single Image Dehazing

The deep convolutional neural networks (CNNs)-based single image dehazin...
research
08/30/2023

Robust Principles: Architectural Design Principles for Adversarially Robust CNNs

Our research aims to unify existing works' diverging opinions on how arc...

Please sign up or login with your details

Forgot password? Click here to reset