Acta Photonica Sinica, Volume. 50, Issue 11, 1110001(2021)
An Infrared Object Detection Method Based on Cross-domain Fusion Network
Figures & Tables(11)
Table 1. Ablation study
View tableView in ArticleTable 1. Ablation study
Modal transfer network Cross-domain fusion networks Soft weight allocation module mAP/% × × × 78 √ × √ 80.3 √ √ × 81.8 √ √ √ 86.4
Table 2. Performance comparison of same network under different datasets
View tableView in ArticleTable 2. Performance comparison of same network under different datasets
Methods Backbone Datasets Precision Recall F1-Score mAP/% Mask R-CNN Resnet101-FPN FLIR 0.640 0.902 0.748 84.0 Pseudo Visible FLIR 0.617 0.921 0.738 83.3
Table 3. Performance comparison of different networks under different datasets
View tableView in ArticleTable 3. Performance comparison of different networks under different datasets
Methods Backbone Datasets AP/% mAP/% FPS Car Person Bicycle Faster R-CNN Resnet101-FPN FLIR 83.7 78.3 74.7 78.9 17 SODA 79.3 78.5 73.4 77.1 YOLO-V4 CSPDarknet-53 FLIR 80 78.8 75.7 78.1 39 SODA 82.7 78.5 70.1 77.1 MMTOD Resnet101-FPN FLIR 86.8 83.1 79.1 83 10 SODA 84.2 81.4 76 80.5 Baseline Resnet101-FPN FLIR 88.3 84.6 79.6 84.2 14 SODA 84.8 83.6 76.5 81.6 Ours Resnet101-FPN FLIR 91.3 88.5 83.2 87.7 16 SODA 88 86.9 80.4 85.1
Tools
Get Citation
Copy Citation Text
Ming ZHAO, Haoran ZHANG. An Infrared Object Detection Method Based on Cross-domain Fusion Network[J]. Acta Photonica Sinica, 2021, 50(11): 1110001
Paper Information
Category:
Received: Apr. 30, 2021
Accepted: Jul. 1, 2021
Published Online: Dec. 2, 2021
The Author Email: ZHANG Haoran (hhhxl_zhr@163.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20