Lightweight 3D DenseNet with improved attention mechanism

Qiaorong Liu; Wanyun Liu

doi:10.1117/12.2658862

23 November 2022 Lightweight 3D DenseNet with improved attention mechanism

Qiaorong Liu, Wanyun Liu

Proceedings Volume 12454, International Symposium on Robotics, Artificial Intelligence, and Information Engineering (RAIIE 2022); 1245427 (2022) https://doi.org/10.1117/12.2658862
Event: International Symposium on Robotics, Artificial Intelligence, and Information Engineering (RAIIE 2022), 2022, Hohhot, China

Abstract

As an important branch in the field of computer vision, the exploration of video understanding is still ongoing, and its important part of video understanding has become one of the research hot spots. Firstly, aiming at the problems of large amount of convolution network parameters and model degradation caused by the deepening of network layers, DDensenet network is proposed, and dynamic grouping convolution module (DGC) is adopted, which greatly reduces the amount of parameters, improves the calculation efficiency and alleviates the problem of gradient disappearance. Secondly, the improved convolutional attention module (S-CBMA) is added, and the adaptive module is added to its spatial attention mechanism, which can distribute the weight of features and weaken the background at the same time. DDensenet network model achieves 95.8% and 79.8% average classification accuracy on Sports8 data set and Olympic16 data set respectively. The results show that the model has faster convergence speed and higher classification performance than C3D, ResNet3d, Two-Stream and other network models.

Citation Download Citation

Qiaorong Liu and Wanyun Liu "Lightweight 3D DenseNet with improved attention mechanism", Proc. SPIE 12454, International Symposium on Robotics, Artificial Intelligence, and Information Engineering (RAIIE 2022), 1245427 (23 November 2022); https://doi.org/10.1117/12.2658862

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
5 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Convolution

Video

Data modeling

3D modeling

Visualization

Detection and tracking algorithms

Feature extraction

Show All Keywords

Keywords/Phrases

Search In:

Publication Years