Yolov4-Sensitive: Feature sensitive multiscale object detection network

Ze-qing Zhou; Zhou-yu Gu

doi:10.1117/12.2673687

22 May 2023 Yolov4-Sensitive: Feature sensitive multiscale object detection network

Ze-qing Zhou, Zhou-yu Gu

Proceedings Volume 12640, International Conference on Internet of Things and Machine Learning (IoTML 2022); 1264010 (2023) https://doi.org/10.1117/12.2673687
Event: International Conference on Internet of Things and Machine Learning (IoTML 2022), 2022, Harbin, China

Abstract

In this paper, we propose the YOLOv4-Sensitive algorithm based on the YOLOv4 algorithm. Firstly, the residual cell structure in the backbone feature extraction network CSPDarknet53 is reconstructed, and a new feature extraction network U-CSPDarknet53 is designed to extract and retain small target feature information at finer granularity. Secondly, MFENet (Multi-receptive Field Extraction Network) network is designed to extract the contextual information of small targets using parallel expansion convolutional branches to alleviate the feature loss problem of SPP network due to pooling operation. Finally, CA attention mechanism is introduced to design a multi-scale feature fusion network CA-PANet to integrate location information into the feature aggregation network and enhance the feature description capability. The algorithm is validated on the PASCAL VOC dataset. the improvement of YOLOv4-Sensitive for small target detection accuracy is more obvious, with an improvement of about 4.02 percentage points. In addition, the inference speed of the algorithm in this paper is 42 frames/second, which improves the small target detection accuracy without losing the inference speed.

Citation Download Citation

Ze-qing Zhou and Zhou-yu Gu "Yolov4-Sensitive: Feature sensitive multiscale object detection network", Proc. SPIE 12640, International Conference on Internet of Things and Machine Learning (IoTML 2022), 1264010 (22 May 2023); https://doi.org/10.1117/12.2673687

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
12 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Small targets

Feature extraction

Target detection

Detection and tracking algorithms

Feature fusion

Convolution

Neural networks

Show All Keywords

Keywords/Phrases

Search In:

Publication Years