Model compression algorithm and model recovery strategy based on channel importance

Zhixiong Zhang; Yun Wu

doi:10.1117/12.3026123

8 April 2024 Model compression algorithm and model recovery strategy based on channel importance

Zhixiong Zhang, Yun Wu

Proceedings Volume 13090, International Conference on Computer Application and Information Security (ICCAIS 2023); 130902G (2024) https://doi.org/10.1117/12.3026123
Event: International Conference on Computer Application and Information Security (ICCAIS 2023), 2023, Wuhan, China

Abstract

In the field of deep learning, object detection algorithms, especially the RetinaNet model based on convolutional neural networks, have garnered considerable attention due to their excellent multi-scale object detection performance. However, improvements in network performance often come at the cost of sacrificing inference speed, particularly when making enhancements to the network structure. This paper focuses on mitigating this negative impact by optimizing the improved RetinaNet model using channel pruning techniques. Initially, we conducted a quantitative analysis to assess the importance of each channel within the network to determine a pruning strategy. Subsequently, the network was finely tuned based on the set pruning rate, effectively reducing the model’s parameter count and computational complexity. It is noteworthy that the pruning process may lead to a decrease in detection performance. To address this, we introduced a novel model recovery method based on channel importance to compensate for the loss in performance. Experimental results indicate that this method significantly improves the model’s inference speed while maintaining a relatively high detection accuracy. The achievements of this research not only provide a new enhancement path for the practicality of the RetinaNet model but also bring a fresh perspective to the field of deep learning model optimization.

(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Zhixiong Zhang and Yun Wu "Model compression algorithm and model recovery strategy based on channel importance", Proc. SPIE 13090, International Conference on Computer Application and Information Security (ICCAIS 2023), 130902G (8 April 2024); https://doi.org/10.1117/12.3026123

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
6 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Data modeling

Matrices

Object detection

Education and training

Performance modeling

Machine learning

Mathematical modeling

Show All Keywords

Keywords/Phrases

Search In:

Publication Years