Journal of Electronic Science and Technology, Volume. 22, Issue 1, 100243(2024)
Benchmarking YOLOv5 models for improved human detection in search and rescue missions
Figures & Tables(14)
Fig. 3. Input and network image resolution: (a) input image resolution and (b) network image resolution.
Fig. 4. Classifier at different IOU thresholds: (a) precision and (b) recall.
Fig. 5. Average precision: (a) precision–to-recall and (b)
Table 1. Network resolution (%) YOLOv5 detection performance.
View tableView in ArticleTable 1. Network resolution (%) YOLOv5 detection performance.
Network resolutiontest AP AP50 320×320 38 78 640×640 60 95 832×832 64 96
Table 2. Hardware machine specification of Google Collaboratory.
View tableView in ArticleTable 2. Hardware machine specification of Google Collaboratory.
Parameter Specification Model name Intel(R) Xeon(R) CPU @ 2.30 GHz CPU MHz 2299.998 Cache size 46080 KB CPU cores 2 RAM 12 GB GPU Nvidia K80 / T4 GPU memory clock 0.82 GHz / 1.59 GHz Max execution time 12 hours Max idle time 90 minutes Model name Intel(R) Xeon(R) CPU @ 2.30 GHz CPU (MHz) 2299.998
Table 3. Comparative results on the SAR dataset.
View tableView in ArticleTable 3. Comparative results on the SAR dataset.
Model Class Images Labels mAP@0.5 mAP @0.75 mAP YOLOv5L All 792 2605 0.969 0.743 0.643 YOLOv4 All 792 2605 0.960 0.710 0.610 YOLOv3 All 792 2605 0.925 0.63 0.902 Faster R-CNN All 792 2605 0.910 0.510 0.500
Table 4. Precision-to-recall ratios for different models.
View tableView in ArticleTable 4. Precision-to-recall ratios for different models.
Model Precision Recall YOLOv5L 0.971 0.932 YOLOv4 0.960 0.910 YOLOv3 0.962 0.892 Faster R-CNN 0.670 0.936
Table 5. Performance comparison of YOLOv5 models on the SAR dataset.
View tableView in ArticleTable 5. Performance comparison of YOLOv5 models on the SAR dataset.
Model Precision Recall mAP YOLOv5s 0.940 0.917 0.933 YOLOv5m 0.775 0.767 0.762 YOLOv5L 0.971 0.932 0.969
Table 6. Performance comparison of YOLOv5 models on the HERIDAL dataset.
View tableView in ArticleTable 6. Performance comparison of YOLOv5 models on the HERIDAL dataset.
Model Precision Recall mAP YOLOv5s 0.753 0.694 0.731 YOLOv5m 0.797 0.812 0.810 YOLOv5L 0.900 0.893 0.864
Table 7. Comparison of YOLOv3, YOLOv4, YOLOv5L and faster R-CNN models on the HERIDAL dataset.
View tableView in ArticleTable 7. Comparison of YOLOv3, YOLOv4, YOLOv5L and faster R-CNN models on the HERIDAL dataset.
Model Precision Recall mAP YOLOv5L 0.900 0.893 0.864 YOLOv4 0.860 0.830 0.783 YOLOv3 0.877 0.780 0.752 Faster R-CNN 0.770 0.890 0.861
Tools
Get Citation
Copy Citation Text
Namat Bachir, Qurban Ali Memon. Benchmarking YOLOv5 models for improved human detection in search and rescue missions[J]. Journal of Electronic Science and Technology, 2024, 22(1): 100243
Paper Information
Category:
Received: Jan. 19, 2023
Accepted: Feb. 4, 2024
Published Online: Jul. 5, 2024
The Author Email: Memon Qurban Ali (qurban.memon@uaeu.ac.ae)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20