Deep learning based multi-branch structure detection network

Pengfei Li

doi:10.1117/12.2686661

10 August 2023 Deep learning based multi-branch structure detection network

Pengfei Li

Author Affiliations +

Proceedings Volume 12759, International Conference on Automation Control, Algorithm, and Intelligent Bionics (ACAIB 2023); 1275925 (2023) https://doi.org/10.1117/12.2686661
Event: 2023 3rd International Conference on Automation Control, Algorithm and Intelligent Bionics (ACAIB 2023), 2023, Xiamen, China

Abstract

Convolutional neural networks perform extremely well in current image classification tasks based on deep learning. In our research experiment, we found a flaw and made a simple but extremely effective improvement. In this article, we used the traditional ResNet and VGG16 as the benchmark networks. By parallelizing multiple networks into a multi-branch structure model, we designed a multi-branch structure classification header, arranged in parallel order into a whole model, and sent the entire image into the detection header of the multi-branch structure. Through different convolution kernels, we obtained the category scores of multiple categories, and output the highest score. Adding a multi-branch architecture as an external module to the target detection network has been experimentally found to improve performance. Our experiments on cifar-10, cifar-100, and ImageNet 2012 datasets have shown that our method can achieve extremely high classification network prediction accuracy under extreme conditions. Under conventional conditions, the accuracy rate also reaches 97%. Applying related modules to target detection networks also has relatively improved performance.

Citation Download Citation

Pengfei Li "Deep learning based multi-branch structure detection network", Proc. SPIE 12759, International Conference on Automation Control, Algorithm, and Intelligent Bionics (ACAIB 2023), 1275925 (10 August 2023); https://doi.org/10.1117/12.2686661

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