TY - JOUR
T1 - A Lightweight Block With Information Flow Enhancement for Convolutional Neural Networks
AU - Bao, Zhiqiang
AU - Yang, Shunzhi
AU - Huang, Zhenhua
AU - Zhou, Meng Chu
AU - Chen, Yunwen
N1 - Funding Information:
This work was supported in part by the National Natural Science Foundation of China under Grant 62172166 and Grant 61772366, in part by the Guangdong Basic and Applied Basic Research Foundation under Grant 2022A1515011380, and in part by the Ministry of Science and Higher Education of the Russian Federation as part of World-class Research Center Program: Advanced Digital Technologies under Contract 075-15-2020-903.
Publisher Copyright:
© 2023 IEEE.
PY - 2023/8/1
Y1 - 2023/8/1
N2 - Convolutional neural networks (CNNs) have demonstrated excellent capability in various visual recognition tasks but impose an excessive computational burden. The latter problem is commonly solved by utilizing lightweight sparse networks. However, such networks have a limited receptive field in a few layers, and the majority of these networks face a severe information barrage due to their sparse structures. Spurred by these deficiencies, this work proposes a Squeeze Convolution block with Information Flow Enhancement (SCIFE), comprising a Divide-and-Squeeze Convolution and an Information Flow Enhancement scheme. The former module constructs a multi-layer structure through multiple squeeze operations to increase the receptive field and reduce computation. The latter replaces the affine transformation with the point convolution and dynamically adjusts the activation function's threshold, enhancing information flow in both channels and layers. Moreover, we reveal that the original affine transformation may harm the network's generalization capability. To overcome this issue, we utilize a point convolution with a zero initial mean. SCIFE can serve as a plug-and-play replacement for vanilla convolution blocks in mainstream CNNs, while extensive experimental results demonstrate that CNNs equipped with SCIFE compress benchmark structures without sacrificing performance, outperforming their competitors.
AB - Convolutional neural networks (CNNs) have demonstrated excellent capability in various visual recognition tasks but impose an excessive computational burden. The latter problem is commonly solved by utilizing lightweight sparse networks. However, such networks have a limited receptive field in a few layers, and the majority of these networks face a severe information barrage due to their sparse structures. Spurred by these deficiencies, this work proposes a Squeeze Convolution block with Information Flow Enhancement (SCIFE), comprising a Divide-and-Squeeze Convolution and an Information Flow Enhancement scheme. The former module constructs a multi-layer structure through multiple squeeze operations to increase the receptive field and reduce computation. The latter replaces the affine transformation with the point convolution and dynamically adjusts the activation function's threshold, enhancing information flow in both channels and layers. Moreover, we reveal that the original affine transformation may harm the network's generalization capability. To overcome this issue, we utilize a point convolution with a zero initial mean. SCIFE can serve as a plug-and-play replacement for vanilla convolution blocks in mainstream CNNs, while extensive experimental results demonstrate that CNNs equipped with SCIFE compress benchmark structures without sacrificing performance, outperforming their competitors.
KW - Convolutional neural network
KW - activation function
KW - affine transformation
KW - information flow
KW - lightweight
UR - http://www.scopus.com/inward/record.url?scp=85147260767&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85147260767&partnerID=8YFLogxK
U2 - 10.1109/TCSVT.2023.3237615
DO - 10.1109/TCSVT.2023.3237615
M3 - Article
AN - SCOPUS:85147260767
SN - 1051-8215
VL - 33
SP - 3570
EP - 3584
JO - IEEE Transactions on Circuits and Systems for Video Technology
JF - IEEE Transactions on Circuits and Systems for Video Technology
IS - 8
ER -