Trident‐YOLO: Improving the precision and speed of mobile device object detection
Abstract This paper introduce an efficient object detection network named Trident‐You Only Look Once (YOLO), which is designed for mobile devices with limited computing power. The new architecture is improved based on YOLO v4‐tiny. The authors redesign the network structure and propose a trident fea...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2022-01-01
|
| Series: | IET Image Processing |
| Subjects: | |
| Online Access: | https://doi.org/10.1049/ipr2.12340 |
| _version_ | 1828104844069568512 |
|---|---|
| author | Guanbo Wang Hongwei Ding, Bo Li Rencan Nie Yifan Zhao |
| author_facet | Guanbo Wang Hongwei Ding, Bo Li Rencan Nie Yifan Zhao |
| author_sort | Guanbo Wang |
| collection | DOAJ |
| description | Abstract This paper introduce an efficient object detection network named Trident‐You Only Look Once (YOLO), which is designed for mobile devices with limited computing power. The new architecture is improved based on YOLO v4‐tiny. The authors redesign the network structure and propose a trident feature pyramid network (Trident‐FPN), which can improve the precision and recall of lightweight object detection. Specifically, Trident‐FPN increases the computational complexity by only a small amount of floating point operations per second (FLOPs) and obtains a multi‐scale feature map of the model, which significantly lightweight object detection performance. To enlarge the receptive field of the network with the fewest FLOPs, this paper redesign the receptive field block (RFB) and spatial pyramid pooling (SPP) layer and propose tinier cross‐stage partial RFBs and smaller cross‐stage partial SPPs. This paper present extensive experiments, and Trident‐YOLO shows strong performance compared to that of other popular models on the PASCAL VOC and MS COCO. On the MS COCO and PASCAL VOC 2007 test sets, the mean average precision (mAP) of Trident‐YOLO improved by 4.5% and 5.0%, respectively. Trident‐YOLO also reduce the network size by more than 54.4% compared to YOLO v4‐tiny. With a 23.7% FLOP reduction, the FPS is improved by 1.9 on an Nvidia Jetson Xavier NX. |
| first_indexed | 2024-04-11T09:53:10Z |
| format | Article |
| id | doaj.art-798566b7e960453faf3e010888559989 |
| institution | Directory Open Access Journal |
| issn | 1751-9659 1751-9667 |
| language | English |
| last_indexed | 2024-04-11T09:53:10Z |
| publishDate | 2022-01-01 |
| publisher | Wiley |
| record_format | Article |
| series | IET Image Processing |
| spelling | doaj.art-798566b7e960453faf3e0108885599892022-12-22T04:30:43ZengWileyIET Image Processing1751-96591751-96672022-01-0116114515710.1049/ipr2.12340Trident‐YOLO: Improving the precision and speed of mobile device object detectionGuanbo Wang0Hongwei Ding,1Bo Li2Rencan Nie3Yifan Zhao4School of Information Yunnan University Kunming, Yunnan Province Kunming ChinaSchool of Information Yunnan University Kunming, Yunnan Province Kunming ChinaSchool of Information Yunnan University Kunming, Yunnan Province Kunming ChinaSchool of Information Yunnan University Kunming, Yunnan Province Kunming ChinaSchool of Information Yunnan University Kunming, Yunnan Province Kunming ChinaAbstract This paper introduce an efficient object detection network named Trident‐You Only Look Once (YOLO), which is designed for mobile devices with limited computing power. The new architecture is improved based on YOLO v4‐tiny. The authors redesign the network structure and propose a trident feature pyramid network (Trident‐FPN), which can improve the precision and recall of lightweight object detection. Specifically, Trident‐FPN increases the computational complexity by only a small amount of floating point operations per second (FLOPs) and obtains a multi‐scale feature map of the model, which significantly lightweight object detection performance. To enlarge the receptive field of the network with the fewest FLOPs, this paper redesign the receptive field block (RFB) and spatial pyramid pooling (SPP) layer and propose tinier cross‐stage partial RFBs and smaller cross‐stage partial SPPs. This paper present extensive experiments, and Trident‐YOLO shows strong performance compared to that of other popular models on the PASCAL VOC and MS COCO. On the MS COCO and PASCAL VOC 2007 test sets, the mean average precision (mAP) of Trident‐YOLO improved by 4.5% and 5.0%, respectively. Trident‐YOLO also reduce the network size by more than 54.4% compared to YOLO v4‐tiny. With a 23.7% FLOP reduction, the FPS is improved by 1.9 on an Nvidia Jetson Xavier NX.https://doi.org/10.1049/ipr2.12340Optical, image and video signal processingComputational complexityComputer vision and image processing techniques |
| spellingShingle | Guanbo Wang Hongwei Ding, Bo Li Rencan Nie Yifan Zhao Trident‐YOLO: Improving the precision and speed of mobile device object detection IET Image Processing Optical, image and video signal processing Computational complexity Computer vision and image processing techniques |
| title | Trident‐YOLO: Improving the precision and speed of mobile device object detection |
| title_full | Trident‐YOLO: Improving the precision and speed of mobile device object detection |
| title_fullStr | Trident‐YOLO: Improving the precision and speed of mobile device object detection |
| title_full_unstemmed | Trident‐YOLO: Improving the precision and speed of mobile device object detection |
| title_short | Trident‐YOLO: Improving the precision and speed of mobile device object detection |
| title_sort | trident yolo improving the precision and speed of mobile device object detection |
| topic | Optical, image and video signal processing Computational complexity Computer vision and image processing techniques |
| url | https://doi.org/10.1049/ipr2.12340 |
| work_keys_str_mv | AT guanbowang tridentyoloimprovingtheprecisionandspeedofmobiledeviceobjectdetection AT hongweiding tridentyoloimprovingtheprecisionandspeedofmobiledeviceobjectdetection AT boli tridentyoloimprovingtheprecisionandspeedofmobiledeviceobjectdetection AT rencannie tridentyoloimprovingtheprecisionandspeedofmobiledeviceobjectdetection AT yifanzhao tridentyoloimprovingtheprecisionandspeedofmobiledeviceobjectdetection |