Simple and effective multi-scale fusion strategy

LiChong Liang; Jing Chen; HuaiYi Xie

doi:10.1117/12.2658839

23 November 2022 Simple and effective multi-scale fusion strategy

LiChong Liang, Jing Chen, HuaiYi Xie

Proceedings Volume 12454, International Symposium on Robotics, Artificial Intelligence, and Information Engineering (RAIIE 2022); 124540A (2022) https://doi.org/10.1117/12.2658839
Event: International Symposium on Robotics, Artificial Intelligence, and Information Engineering (RAIIE 2022), 2022, Hohhot, China

Abstract

for ground truths in different sizes can effectively improve the detection performance. This idea can be transplanted to classification tasks. Based on the inspiration of object detection, we re-examine the multi-scale information, simplify the steps of multi-scale fusion, and further propose two strategies for multi-scale information fusion based on weight redistribution. On classification tasks, our method maintains a certain degree of competitiveness with other finely designed fusion strategies, and improves the performance of the baseline by 1-7% on 6 datasets. Open code in: https://github.com/Clichong/StageCNN

Citation Download Citation

LiChong Liang, Jing Chen, and HuaiYi Xie "Simple and effective multi-scale fusion strategy", Proc. SPIE 12454, International Symposium on Robotics, Artificial Intelligence, and Information Engineering (RAIIE 2022), 124540A (23 November 2022); https://doi.org/10.1117/12.2658839

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