Bowen Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Bowen Wang has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
- Chongfeng Wei
James Watt School of Engineering, University of Glasgow, Glasgow, U.K.
1 shared publication - Hao Liu
The Beijing Intelligent Transport Development Center, Beijing 100053, China
1 shared publication - Hongcheng Huang
School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai, China
1 shared publication - Ng Chin Long
School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai, China
1 shared publication - Shuai Wang
School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
1 shared publication - Yafei Wang
Shanghai Jiao Tong University
1 shared publication
Latest Publications
- Open Access
Dynamic Occlusion–Predictive Neural Network for Robust Roadside Multi-Vehicle Tracking
Despite their extended detection ranges and superior precision compared with onboard sensors, roadside perception systems suffer from severe occlusion artifacts in complex traffic, causing significant tracking failures and ID switches. To address this, we propose a novel Dynamic Occlusion–Predictive Neural Network tailored to challenging roadside environments. First, we introduce a Transformer-based Dynamic Occlusion State Predictor to explicitly model the temporal evolution of occlusion. Unlike traditional tracking methods, this module continuously forecasts future occlusion ratios for each target by analyzing historical occlusion patterns. Critically, these predictions are integrated into the tracking framework as dynamic weighting factors in the loss function, enabling the model to adaptively penalize tracking errors based on the predicted occlusion severity and significantly enhancing robustness against dynamic occlusion scenarios. Second, leveraging the predicted occlusion states, we propose a GNN-based Spatial Reasoning Module to address trajectory fragmentation. This module constructs a heterogeneous graph integrating road occupancy information and neighboring vehicle poses to infer the existence and motion patterns of targets within occluded regions. By analyzing scene-level physical constraints, it generates motion predictions for invisible targets and links these inferred states to fragmented trajectories, ensuring temporally continuous tracking even during prolonged visual occlusions. Experiments on the DAIR-V2X and our self-collected roadside dataset show that our framework outperforms state-of-the-art methods in precision and robustness, achieving a 5.1% MOTA gain over the best baseline. This advantage peaks under high occlusion, where preserving ID continuity and minimizing failures validates its efficacy for real-world roadside multi-target tracking.
Open AccessDynamic Occlusion–Predictive Neural Network for Robust Roadside Multi-Vehicle Tracking
Despite their extended detection ranges and superior precision compared with onboard sensors, roadside perception systems suffer from severe occlusion artifacts in complex traffic, causing significant tracking failures and ID switches. To address this, we propose a novel Dynamic Occlusion–Predictive Neural Network tailored to challenging roadside environments. First, we introduce a Transformer-based Dynamic Occlusion State Predictor to explicitly model the temporal evolution of occlusion. Unlike traditional tracking methods, this module continuously forecasts future occlusion ratios for each target by analyzing historical occlusion patterns. Critically, these predictions are integrated into the tracking framework as dynamic weighting factors in the loss function, enabling the model to adaptively penalize tracking errors based on the predicted occlusion severity and significantly enhancing robustness against dynamic occlusion scenarios. Second, leveraging the predicted occlusion states, we propose a GNN-based Spatial Reasoning Module to address trajectory fragmentation. This module constructs a heterogeneous graph integrating road occupancy information and neighboring vehicle poses to infer the existence and motion patterns of targets within occluded regions. By analyzing scene-level physical constraints, it generates motion predictions for invisible targets and links these inferred states to fragmented trajectories, ensuring temporally continuous tracking even during prolonged visual occlusions. Experiments on the DAIR-V2X and our self-collected roadside dataset show that our framework outperforms state-of-the-art methods in precision and robustness, achieving a 5.1% MOTA gain over the best baseline. This advantage peaks under high occlusion, where preserving ID continuity and minimizing failures validates its efficacy for real-world roadside multi-target tracking.
- Open Access
- 1 September 2025
- Published by IOP Publishing in Journal of Physics: Conference Series
- Vol. 3125 (1)
SAMTrans: Leveraging the Segment Anything Model as a Bridge for Vehicle-Infrastructure Cross-domain Transfer to Enhance Roadside 3D Object Detection
Roadside LiDAR-based perception has the potential to provide comprehensive and high-precision sensing information, which is of critical importance for safety and efficiency in intelligent transportation systems. However, when deploying 3D roadside perception methods in new traffic scenarios, data acquisition and annotation are labor-intensive and costly. Unsupervised Domain Adaptation (UDA) can extract knowledge from existing data and transfer it to unknown domains for self-supervised model training, making it a promising approach to reduce the need for labeled roadside perception data. However, the effectiveness of UDA is constrained by the generalization capability of the pre-trained model from source domain. Foundational pre-trained large models such as SAM hold the potential for achieving cross-domain generalization across diverse scenarios. Inspired by this sight, this paper proposes a Segment Anything Model Transfer (SAMTrans) approach to enhance model performance. Firstly, we design a Bird’s-Eye View (BEV) pseudo-image generation method to enable Segment Anything Model (SAM) to interpret point cloud data. Additionally, we introduce a pseudo label fusion strategy based on confidence scores and Gaussian kernel density estimation. Evaluated on popular 3D roadside benchmarks, the proposed method demonstrates strong competitiveness in improving upon state-of-the-art performance. For instance, when integrated with the ST3D framework on the nuScenes to V2X-Seq task, our method boosts the 3D Average Precision (AP 3D ) from 30.1% to 40.6% under the Easy difficulty setting.
- 1 September 2025
- Published by IOP Publishing in Journal of Physics: Conference Series
- Vol. 3125 (1)
Open AccessSAMTrans: Leveraging the Segment Anything Model as a Bridge for Vehicle-Infrastructure Cross-domain Transfer to Enhance Roadside 3D Object Detection
Roadside LiDAR-based perception has the potential to provide comprehensive and high-precision sensing information, which is of critical importance for safety and efficiency in intelligent transportation systems. However, when deploying 3D roadside perception methods in new traffic scenarios, data acquisition and annotation are labor-intensive and costly. Unsupervised Domain Adaptation (UDA) can extract knowledge from existing data and transfer it to unknown domains for self-supervised model training, making it a promising approach to reduce the need for labeled roadside perception data. However, the effectiveness of UDA is constrained by the generalization capability of the pre-trained model from source domain. Foundational pre-trained large models such as SAM hold the potential for achieving cross-domain generalization across diverse scenarios. Inspired by this sight, this paper proposes a Segment Anything Model Transfer (SAMTrans) approach to enhance model performance. Firstly, we design a Bird’s-Eye View (BEV) pseudo-image generation method to enable Segment Anything Model (SAM) to interpret point cloud data. Additionally, we introduce a pseudo label fusion strategy based on confidence scores and Gaussian kernel density estimation. Evaluated on popular 3D roadside benchmarks, the proposed method demonstrates strong competitiveness in improving upon state-of-the-art performance. For instance, when integrated with the ST3D framework on the nuScenes to V2X-Seq task, our method boosts the 3D Average Precision (AP 3D ) from 30.1% to 40.6% under the Easy difficulty setting.