Higher efficient YOLOv7: a one-stage method for non-salient object detection

被引:4
作者
Dong, Chengang [1 ]
Tang, Yuhao [1 ]
Zhang, Liyan [1 ]
机构
[1] Nanjing Univ Aeronaut & Astronaut, Nanjing 210000, Peoples R China
基金
中国国家自然科学基金;
关键词
Non-salient; Object Detection; Attention Mechanisms; YOLOv7;
D O I
10.1007/s11042-023-17185-w
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Compared to the remarkable progress within the discipline of object detection in recent years, real-time detection of non-salient objects remains a challenging research task. However, most existing detection methods fail to adequately extract the global features of targets, leading to suboptimal performance when dealing with non-salient objects. In this paper, we propose a unified framework called Higher efficient (He)-YOLOv7 to enhance the detection capability of YOLOv7 for non-salient objects.Firstly, we introduce an refined Squeeze and Excitation Network (SENet) to dynamically adjust the weights of feature channels, thereby enhancing the model's perception of non-salient objects. Secondly, we design an Angle Intersection over Union (AIoU) loss function that considers relative positional information, optimizing the widely used Complete Intersection over Union (CIoU) loss function in YOLOv7. This significantly accelerates the model's convergence. Moreover, He-YOLOv7 adopts a blended data augmentation strategy to simulate occlusion among objects, further improving the model's ability to filter out noise information and enhancing its robustness. Comparison of experimental results demonstrates a significant improvement of 2.4% mean Average Precision (mAP) on the Microsoft Common Objects in Context (MS COCO) dataset and a notable enhancement of 1.2% mAP on the PASCAL VOC dataset. Simultaneously, our approach demonstrates comparable performance to state-of-the-art real-time object detection methods.
引用
收藏
页码:42257 / 42283
页数:27
相关论文
共 61 条
[1]  
Ali H, 2019, INT C INF SCI COMM T, P1
[2]  
Bochkovskiy A, 2020, Arxiv, DOI arXiv:2004.10934
[3]  
Breger A, 2019, Arxiv, DOI arXiv:1910.13422
[4]   Cascade R-CNN: Delving into High Quality Object Detection [J].
Cai, Zhaowei ;
Vasconcelos, Nuno .
2018 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2018, :6154-6162
[5]   Target location detection of mobile robots based on R-FCN deep convolutional neural network [J].
Cen, Hua .
INTERNATIONAL JOURNAL OF SYSTEM ASSURANCE ENGINEERING AND MANAGEMENT, 2023, 14 (02) :728-737
[6]  
Chen K, 2018, P IEEE C COMP VIS PA, P6298
[7]  
Chen Yu, 2020, Advances in Neural Information Processing Systems, V33
[8]   RetinaNet With Difference Channel Attention and Adaptively Spatial Feature Fusion for Steel Surface Defect Detection [J].
Cheng, Xun ;
Yu, Jianbo .
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, 2021, 70 (70)
[9]   FS-COCO: Towards Understanding of Freehand Sketches of Common Objects in Context [J].
Chowdhury, Pinaki Nath ;
Sain, Aneeshan ;
Bhunia, Ayan Kumar ;
Xiang, Tao ;
Gryaditskaya, Yulia ;
Song, Yi-Zhe .
COMPUTER VISION, ECCV 2022, PT VIII, 2022, 13668 :253-270
[10]  
Christlein Vincent, 2019, 2019 International Conference on Document Analysis and Recognition (ICDAR). Proceedings, P1090, DOI 10.1109/ICDAR.2019.00177