Optics and Precision Engineering, Volume. 32, Issue 12, 1929(2024)
Few-shot warhead fragment group object detection based on feature reassembly and attention
Figures & Tables(14)
Fig. 4. Framework for training object detection model using MAML method
Fig. 7. Comparison of precision and recall metrics for different models
Fig. 8. Comparison of average precision metrics for different models
Fig. 9. Comparison of YOLOv5s and this paper model detection effect scene 1
Fig. 10. Comparison of YOLOv5s and this paper model detection effect scene 2
Table 1. Experimental environment
View tableView in ArticleTable 1. Experimental environment
Project Environment CPU Intel（R） Xeon（R） Silver 4215R CPU @ 3.20 GHz 3.19 GHz Memory 128 GB Operating System Ubuntu 20.04 GPU 4 NVIDIA RTX3090 Python version Python3.9 Pytorch version PyTorch1.10 Object Detection Framework YOLOv5s-FD GPU Acceleration Library CUDA cuDNN
Table 2. Comparative experiment on the effectiveness of MAML meta learning training methods
View tableView in ArticleTable 2. Comparative experiment on the effectiveness of MAML meta learning training methods
Training Method P R mAP_0.5 mAP_0.5∶0.95 Traditional 0.862 0.793 0.845 0.546 MAML 0.905 0.854 0.882 0.611
Table 3. Ablation experimental results
View tableView in ArticleTable 3. Ablation experimental results
YOLOv5s SODL CARAFE CA P R mAP_0.5 mAP_0.5∶0.95 Size
/MB
Params
/M
GFLOPs
/G
FPS（frames/s） 1 √ 0.834 0.775 0.807 0.542 14.4 7.02 15.9 127 2 √ √ 0.869 0.813 0.837 0.576 14.8 7.17 18.3 114 3 √ √ √ 0.878 0.826 0.851 0.588 15.1 7.32 19.4 104 4 √ √ √ √ 0.905 0.854 0.882 0.611 15.6 7.94 20.1 95
Table 4. Performance comparison of different algorithm models on Fragment data set
View tableView in ArticleTable 4. Performance comparison of different algorithm models on Fragment data set
Model P R mAP_0.5 mAP_0.5：0.95 Size
/MB
Params
/M
GFLOPs
/G
FPS（frames/s） Faster R-CNN 0.699 0.568 0.598 - 108.9 136.60 402.3 29 SSD 0.630 0.495 0.518 - 100.3 24.40 274.1 81 YOLOv5s 0.834 0.775 0.807 0.542 14.4 7.02 15.9 127 YOLOv8s 0.826 0.789 0.817 0.581 22.5 11.13 28.6 112 YOLOv5s-FD（ours） 0.905 0.854 0.882 0.611 15.6 7.94 20.1 95
Tools
Get Citation
Copy Citation Text
Meng HE, Jiangpeng WU, Chao LIANG, Pengyu HU, Yuan REN, Xuan HE, Qianghui LIU. Few-shot warhead fragment group object detection based on feature reassembly and attention[J]. Optics and Precision Engineering, 2024, 32(12): 1929
Paper Information
Category:
Received: Mar. 9, 2024
Accepted: --
Published Online: Aug. 28, 2024
The Author Email: WU Jiangpeng (wjp62795@126.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20