Research on Image Recognition Based on Improved YOLOv5s Network Model Algorithm

In order to achieve rapid classification and extraction of pig target features. Propose an improved pig image object detection method based on YOLOv5s, using YOLOv5s adaptive anchor box calculation to obtain anchor boxes suitable for self built datasets; Combining convolutional block attention modul...

Full description

Saved in:
Bibliographic Details
Published in2024 4th International Conference on Neural Networks, Information and Communication (NNICE) pp. 658 - 662
Main Authors Xie, Keying, Li, Haoyuan, Chen, Xiaodan, Zheng, Weichao, Gan, Jing, Chen, Xinxin
Format Conference Proceeding
LanguageEnglish
Published IEEE 19.01.2024
Subjects
Adaptive systems
Analytical models
Artificial neural networks
Bidirectional control
Deep learning
Image processing
Image recognition
Network structure
Object detection
Training
YOLO
YOLOv5s
Online AccessGet full text

Cover

More Information
Summary:In order to achieve rapid classification and extraction of pig target features. Propose an improved pig image object detection method based on YOLOv5s, using YOLOv5s adaptive anchor box calculation to obtain anchor boxes suitable for self built datasets; Combining convolutional block attention module to extract feature backbone network, improving model detection performance and reducing parameters. The Mosaic data augmentation method was used to enhance the image richness of the dataset and enhance the feature fusion ability of the network model. The improved algorithm was compared with the original YOLOv5s algorithm on a self-made dataset for ablation experiments. The results showed that the mean mAP of the improved algorithm reached 0.952, which showed a significant improvement in detection accuracy compared to other mainstream algorithms.
DOI:10.1109/NNICE61279.2024.10498747