Low-Cost Object Detection Models for Traffic Control Devices through Domain Adaption of Geographical Regions
Automated inspection systems utilizing computer vision technology are effective in managing traffic control devices (TCDs); however, they face challenges due to the limited availability of training datasets and the difficulty in generating new datasets. To address this, our study establishes a bench...
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-05-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/15/10/2584 |
| _version_ | 1797598469300420608 |
|---|---|
| author | Dahyun Oh Kyubyung Kang Sungchul Seo Jinwu Xiao Kyochul Jang Kibum Kim Hyungkeun Park Jeonghun Won |
| author_facet | Dahyun Oh Kyubyung Kang Sungchul Seo Jinwu Xiao Kyochul Jang Kibum Kim Hyungkeun Park Jeonghun Won |
| author_sort | Dahyun Oh |
| collection | DOAJ |
| description | Automated inspection systems utilizing computer vision technology are effective in managing traffic control devices (TCDs); however, they face challenges due to the limited availability of training datasets and the difficulty in generating new datasets. To address this, our study establishes a benchmark for cost-effective model training methods that achieve the desired accuracy using data from related domains and YOLOv5, a one-stage object detector known for its high accuracy and speed. In this study, three model cases were developed using distinct training approaches: (1) training with COCO-based pre-trained weights, (2) training with pre-trained weights from the source domain, and (3) training with a synthesized dataset mixed with source and target domains. Upon comparing these model cases, this study found that directly applying source domain data to the target domain is unfeasible, and a small amount of target domain data is necessary for optimal performance. A model trained with fine-tuning-based domain adaptation using pre-trained weights from the source domain and minimal target data, proved to be the most resource-efficient approach. These results contribute valuable guidance for practitioners aiming to develop TCD models with limited data, enabling them to build optimal models while conserving resources. |
| first_indexed | 2024-03-11T03:21:35Z |
| format | Article |
| id | doaj.art-972b6d78ce2e4a36b467e1da330a150e |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-11T03:21:35Z |
| publishDate | 2023-05-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-972b6d78ce2e4a36b467e1da330a150e2023-11-18T03:07:14ZengMDPI AGRemote Sensing2072-42922023-05-011510258410.3390/rs15102584Low-Cost Object Detection Models for Traffic Control Devices through Domain Adaption of Geographical RegionsDahyun Oh0Kyubyung Kang1Sungchul Seo2Jinwu Xiao3Kyochul Jang4Kibum Kim5Hyungkeun Park6Jeonghun Won7Department of Civil Engineering, Chungbuk National University, Cheongju-si 28644, Republic of KoreaSchool of Construction Management Technology, Purdue University, West Lafayette, IN 47907, USADepartment of Civil Engineering, Chungbuk National University, Cheongju-si 28644, Republic of KoreaSchool of Construction Management Technology, Purdue University, West Lafayette, IN 47907, USADepartment of Computer Science, Purdue University, West Lafayette, IN 47907, USADivision of Construction Engineering and Management, Purdue University, West Lafayette, IN 47907, USADepartment of Civil Engineering, Chungbuk National University, Cheongju-si 28644, Republic of KoreaDepartment of Safety Engineering, Chungbuk National University, Cheongju-si 28644, Republic of KoreaAutomated inspection systems utilizing computer vision technology are effective in managing traffic control devices (TCDs); however, they face challenges due to the limited availability of training datasets and the difficulty in generating new datasets. To address this, our study establishes a benchmark for cost-effective model training methods that achieve the desired accuracy using data from related domains and YOLOv5, a one-stage object detector known for its high accuracy and speed. In this study, three model cases were developed using distinct training approaches: (1) training with COCO-based pre-trained weights, (2) training with pre-trained weights from the source domain, and (3) training with a synthesized dataset mixed with source and target domains. Upon comparing these model cases, this study found that directly applying source domain data to the target domain is unfeasible, and a small amount of target domain data is necessary for optimal performance. A model trained with fine-tuning-based domain adaptation using pre-trained weights from the source domain and minimal target data, proved to be the most resource-efficient approach. These results contribute valuable guidance for practitioners aiming to develop TCD models with limited data, enabling them to build optimal models while conserving resources.https://www.mdpi.com/2072-4292/15/10/2584domain adaptationlow-cost object detectiontraffic control devices (TCDs)training dataset benchmarkYOLOv5 |
| spellingShingle | Dahyun Oh Kyubyung Kang Sungchul Seo Jinwu Xiao Kyochul Jang Kibum Kim Hyungkeun Park Jeonghun Won Low-Cost Object Detection Models for Traffic Control Devices through Domain Adaption of Geographical Regions Remote Sensing domain adaptation low-cost object detection traffic control devices (TCDs) training dataset benchmark YOLOv5 |
| title | Low-Cost Object Detection Models for Traffic Control Devices through Domain Adaption of Geographical Regions |
| title_full | Low-Cost Object Detection Models for Traffic Control Devices through Domain Adaption of Geographical Regions |
| title_fullStr | Low-Cost Object Detection Models for Traffic Control Devices through Domain Adaption of Geographical Regions |
| title_full_unstemmed | Low-Cost Object Detection Models for Traffic Control Devices through Domain Adaption of Geographical Regions |
| title_short | Low-Cost Object Detection Models for Traffic Control Devices through Domain Adaption of Geographical Regions |
| title_sort | low cost object detection models for traffic control devices through domain adaption of geographical regions |
| topic | domain adaptation low-cost object detection traffic control devices (TCDs) training dataset benchmark YOLOv5 |
| url | https://www.mdpi.com/2072-4292/15/10/2584 |
| work_keys_str_mv | AT dahyunoh lowcostobjectdetectionmodelsfortrafficcontroldevicesthroughdomainadaptionofgeographicalregions AT kyubyungkang lowcostobjectdetectionmodelsfortrafficcontroldevicesthroughdomainadaptionofgeographicalregions AT sungchulseo lowcostobjectdetectionmodelsfortrafficcontroldevicesthroughdomainadaptionofgeographicalregions AT jinwuxiao lowcostobjectdetectionmodelsfortrafficcontroldevicesthroughdomainadaptionofgeographicalregions AT kyochuljang lowcostobjectdetectionmodelsfortrafficcontroldevicesthroughdomainadaptionofgeographicalregions AT kibumkim lowcostobjectdetectionmodelsfortrafficcontroldevicesthroughdomainadaptionofgeographicalregions AT hyungkeunpark lowcostobjectdetectionmodelsfortrafficcontroldevicesthroughdomainadaptionofgeographicalregions AT jeonghunwon lowcostobjectdetectionmodelsfortrafficcontroldevicesthroughdomainadaptionofgeographicalregions |