Improved YOLOv3 Integrating SENet and Optimized GIoU Loss for Occluded Pedestrian Detection
Occluded pedestrian detection faces huge challenges. False positives and false negatives in crowd occlusion scenes will reduce the accuracy of occluded pedestrian detection. To overcome this problem, we proposed an improved you-only-look-once version 3 (YOLOv3) based on squeeze-and-excitation networ...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-11-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/23/22/9089 |
| _version_ | 1797457845519646720 |
|---|---|
| author | Qiangbo Zhang Yunxiang Liu Yu Zhang Ming Zong Jianlin Zhu |
| author_facet | Qiangbo Zhang Yunxiang Liu Yu Zhang Ming Zong Jianlin Zhu |
| author_sort | Qiangbo Zhang |
| collection | DOAJ |
| description | Occluded pedestrian detection faces huge challenges. False positives and false negatives in crowd occlusion scenes will reduce the accuracy of occluded pedestrian detection. To overcome this problem, we proposed an improved you-only-look-once version 3 (YOLOv3) based on squeeze-and-excitation networks (SENet) and optimized generalized intersection over union (GIoU) loss for occluded pedestrian detection, namely YOLOv3-Occlusion (YOLOv3-Occ). The proposed network model considered incorporating squeeze-and-excitation networks (SENet) into YOLOv3, which assigned greater weights to the features of unobstructed parts of pedestrians to solve the problem of feature extraction against unsheltered parts. For the loss function, a new generalized intersection over union<sub>intersection over groundtruth</sub> (GIoU<sub>IoG</sub>) loss was developed to ensure the areas of predicted frames of pedestrian invariant based on the GIoU loss, which tackled the problem of inaccurate positioning of pedestrians. The proposed method, YOLOv3-Occ, was validated on the CityPersons and COCO2014 datasets. Experimental results show the proposed method could obtain 1.2% MR<sup>−2</sup> gains on the CityPersons dataset and 0.7% mAP@50 improvements on the COCO2014 dataset. |
| first_indexed | 2024-03-09T16:28:36Z |
| format | Article |
| id | doaj.art-9af13fff337f4c569d81eaf153e513fd |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-09T16:28:36Z |
| publishDate | 2023-11-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-9af13fff337f4c569d81eaf153e513fd2023-11-24T15:05:17ZengMDPI AGSensors1424-82202023-11-012322908910.3390/s23229089Improved YOLOv3 Integrating SENet and Optimized GIoU Loss for Occluded Pedestrian DetectionQiangbo Zhang0Yunxiang Liu1Yu Zhang2Ming Zong3Jianlin Zhu4School of Computer Science and Information Engineering, Shanghai Institute of Technology, Shanghai 201418, ChinaSchool of Computer Science and Information Engineering, Shanghai Institute of Technology, Shanghai 201418, ChinaSchool of Computer Science and Information Engineering, Shanghai Institute of Technology, Shanghai 201418, ChinaSchool of Computer Science and Information Engineering, Shanghai Institute of Technology, Shanghai 201418, ChinaSchool of Computer Science and Information Engineering, Shanghai Institute of Technology, Shanghai 201418, ChinaOccluded pedestrian detection faces huge challenges. False positives and false negatives in crowd occlusion scenes will reduce the accuracy of occluded pedestrian detection. To overcome this problem, we proposed an improved you-only-look-once version 3 (YOLOv3) based on squeeze-and-excitation networks (SENet) and optimized generalized intersection over union (GIoU) loss for occluded pedestrian detection, namely YOLOv3-Occlusion (YOLOv3-Occ). The proposed network model considered incorporating squeeze-and-excitation networks (SENet) into YOLOv3, which assigned greater weights to the features of unobstructed parts of pedestrians to solve the problem of feature extraction against unsheltered parts. For the loss function, a new generalized intersection over union<sub>intersection over groundtruth</sub> (GIoU<sub>IoG</sub>) loss was developed to ensure the areas of predicted frames of pedestrian invariant based on the GIoU loss, which tackled the problem of inaccurate positioning of pedestrians. The proposed method, YOLOv3-Occ, was validated on the CityPersons and COCO2014 datasets. Experimental results show the proposed method could obtain 1.2% MR<sup>−2</sup> gains on the CityPersons dataset and 0.7% mAP@50 improvements on the COCO2014 dataset.https://www.mdpi.com/1424-8220/23/22/9089occluded pedestrian detectionfalse positivesfalse negativesloss function |
| spellingShingle | Qiangbo Zhang Yunxiang Liu Yu Zhang Ming Zong Jianlin Zhu Improved YOLOv3 Integrating SENet and Optimized GIoU Loss for Occluded Pedestrian Detection Sensors occluded pedestrian detection false positives false negatives loss function |
| title | Improved YOLOv3 Integrating SENet and Optimized GIoU Loss for Occluded Pedestrian Detection |
| title_full | Improved YOLOv3 Integrating SENet and Optimized GIoU Loss for Occluded Pedestrian Detection |
| title_fullStr | Improved YOLOv3 Integrating SENet and Optimized GIoU Loss for Occluded Pedestrian Detection |
| title_full_unstemmed | Improved YOLOv3 Integrating SENet and Optimized GIoU Loss for Occluded Pedestrian Detection |
| title_short | Improved YOLOv3 Integrating SENet and Optimized GIoU Loss for Occluded Pedestrian Detection |
| title_sort | improved yolov3 integrating senet and optimized giou loss for occluded pedestrian detection |
| topic | occluded pedestrian detection false positives false negatives loss function |
| url | https://www.mdpi.com/1424-8220/23/22/9089 |
| work_keys_str_mv | AT qiangbozhang improvedyolov3integratingsenetandoptimizedgioulossforoccludedpedestriandetection AT yunxiangliu improvedyolov3integratingsenetandoptimizedgioulossforoccludedpedestriandetection AT yuzhang improvedyolov3integratingsenetandoptimizedgioulossforoccludedpedestriandetection AT mingzong improvedyolov3integratingsenetandoptimizedgioulossforoccludedpedestriandetection AT jianlinzhu improvedyolov3integratingsenetandoptimizedgioulossforoccludedpedestriandetection |