An Improved YOLOX Model and Domain Transfer Strategy for Nighttime Pedestrian and Vehicle Detection
Aimed at the vehicle/pedestrian visual sensing task under low-light conditions and the problems of small, dense objects and line-of-sight occlusion, a nighttime vehicle/pedestrian detection method was proposed. First, a vehicle/pedestrian detection algorithm was designed based on You Only Look Once...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2022-12-01
|
Series: | Applied Sciences |
Subjects: | |
Online Access: | https://www.mdpi.com/2076-3417/12/23/12476 |
_version_ | 1797463542856679424 |
---|---|
author | Kefu Yi Kai Luo Tuo Chen Rongdong Hu |
author_facet | Kefu Yi Kai Luo Tuo Chen Rongdong Hu |
author_sort | Kefu Yi |
collection | DOAJ |
description | Aimed at the vehicle/pedestrian visual sensing task under low-light conditions and the problems of small, dense objects and line-of-sight occlusion, a nighttime vehicle/pedestrian detection method was proposed. First, a vehicle/pedestrian detection algorithm was designed based on You Only Look Once X (YOLOX). The model structure was re-parameterized and lightened, and a coordinate-based attention mechanism was introduced into the backbone network to enhance the feature extraction efficiency of vehicle/pedestrian targets. A feature-scale fusion detection branch was added to the feature pyramid, while a loss function was designed, which combines Complete Intersection Over Union (CIoU) for target localization and Varifocal Loss for confidence prediction to improve the feature extraction ability for small, dense, and low-illumination targets. In addition, in order to further improve the detection accuracy of the algorithm under low-light conditions, a training strategy based on data domain transfer was proposed, which fuses the larger-scale daylight dataset with the smaller-scale nighttime dataset after low-illumination degrading. After low-light enhancement, training and testing were performed accordingly. The experimental results show that, compared with the original YOLOX model, the improved algorithm trained by the proposed data domain transfer strategy achieved better performance, and the mean Average Precision (mAP) increased by 5.9% to 82.4%. This research provided effective technical support for autonomous driving safety at night. |
first_indexed | 2024-03-09T17:52:13Z |
format | Article |
id | doaj.art-53256e10613f458f8e55079c0eb145d9 |
institution | Directory Open Access Journal |
issn | 2076-3417 |
language | English |
last_indexed | 2024-03-09T17:52:13Z |
publishDate | 2022-12-01 |
publisher | MDPI AG |
record_format | Article |
series | Applied Sciences |
spelling | doaj.art-53256e10613f458f8e55079c0eb145d92023-11-24T10:37:08ZengMDPI AGApplied Sciences2076-34172022-12-0112231247610.3390/app122312476An Improved YOLOX Model and Domain Transfer Strategy for Nighttime Pedestrian and Vehicle DetectionKefu Yi0Kai Luo1Tuo Chen2Rongdong Hu3College of Automotive and Mechanical Engineering, Changsha University of Science and Technology, Changsha 410114, ChinaCollege of Automotive and Mechanical Engineering, Changsha University of Science and Technology, Changsha 410114, ChinaCollege of Automotive and Mechanical Engineering, Changsha University of Science and Technology, Changsha 410114, ChinaChangsha Intelligent Driving Institute Ltd., Changsha 410208, ChinaAimed at the vehicle/pedestrian visual sensing task under low-light conditions and the problems of small, dense objects and line-of-sight occlusion, a nighttime vehicle/pedestrian detection method was proposed. First, a vehicle/pedestrian detection algorithm was designed based on You Only Look Once X (YOLOX). The model structure was re-parameterized and lightened, and a coordinate-based attention mechanism was introduced into the backbone network to enhance the feature extraction efficiency of vehicle/pedestrian targets. A feature-scale fusion detection branch was added to the feature pyramid, while a loss function was designed, which combines Complete Intersection Over Union (CIoU) for target localization and Varifocal Loss for confidence prediction to improve the feature extraction ability for small, dense, and low-illumination targets. In addition, in order to further improve the detection accuracy of the algorithm under low-light conditions, a training strategy based on data domain transfer was proposed, which fuses the larger-scale daylight dataset with the smaller-scale nighttime dataset after low-illumination degrading. After low-light enhancement, training and testing were performed accordingly. The experimental results show that, compared with the original YOLOX model, the improved algorithm trained by the proposed data domain transfer strategy achieved better performance, and the mean Average Precision (mAP) increased by 5.9% to 82.4%. This research provided effective technical support for autonomous driving safety at night.https://www.mdpi.com/2076-3417/12/23/12476vehicle and pedestrian detectionYOLOXattention mechanismdata domain transfernight object detection |
spellingShingle | Kefu Yi Kai Luo Tuo Chen Rongdong Hu An Improved YOLOX Model and Domain Transfer Strategy for Nighttime Pedestrian and Vehicle Detection Applied Sciences vehicle and pedestrian detection YOLOX attention mechanism data domain transfer night object detection |
title | An Improved YOLOX Model and Domain Transfer Strategy for Nighttime Pedestrian and Vehicle Detection |
title_full | An Improved YOLOX Model and Domain Transfer Strategy for Nighttime Pedestrian and Vehicle Detection |
title_fullStr | An Improved YOLOX Model and Domain Transfer Strategy for Nighttime Pedestrian and Vehicle Detection |
title_full_unstemmed | An Improved YOLOX Model and Domain Transfer Strategy for Nighttime Pedestrian and Vehicle Detection |
title_short | An Improved YOLOX Model and Domain Transfer Strategy for Nighttime Pedestrian and Vehicle Detection |
title_sort | improved yolox model and domain transfer strategy for nighttime pedestrian and vehicle detection |
topic | vehicle and pedestrian detection YOLOX attention mechanism data domain transfer night object detection |
url | https://www.mdpi.com/2076-3417/12/23/12476 |
work_keys_str_mv | AT kefuyi animprovedyoloxmodelanddomaintransferstrategyfornighttimepedestrianandvehicledetection AT kailuo animprovedyoloxmodelanddomaintransferstrategyfornighttimepedestrianandvehicledetection AT tuochen animprovedyoloxmodelanddomaintransferstrategyfornighttimepedestrianandvehicledetection AT rongdonghu animprovedyoloxmodelanddomaintransferstrategyfornighttimepedestrianandvehicledetection AT kefuyi improvedyoloxmodelanddomaintransferstrategyfornighttimepedestrianandvehicledetection AT kailuo improvedyoloxmodelanddomaintransferstrategyfornighttimepedestrianandvehicledetection AT tuochen improvedyoloxmodelanddomaintransferstrategyfornighttimepedestrianandvehicledetection AT rongdonghu improvedyoloxmodelanddomaintransferstrategyfornighttimepedestrianandvehicledetection |