Feiyang Jia

Feiyang Jia

Auto-generated profile, based on publicly available information

Feiyang Jia

School of Computer Science and Technology, Beijing Key Laboratory of Traffic Data Mining and Embodied Intelligence, Beijing Jiaotong University, Beijing, China

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
29 October 2024
Verified
Book Chapter
Closed access
PDF file
ZSZiying Song
Ziying Song
LYLei Yang
Lei Yang
SXShaoqing Xu
Shaoqing Xu
LLLin Liu
Lin Liu
DXDongyang Xu
Dongyang Xu
LWLi Wang
Li Wang

3 November 2025 in IEEE Transactions on Circuits and Systems for Video Technology
Verified
Journal Article
Closed access
FJFeiyang Jia
Feiyang Jia
CJCaiyan Jia
Caiyan Jia
ALAilin Liu
Ailin Liu
SXShaoqing Xu
Shaoqing Xu
QXqiming Xia
qiming Xia
ZSZiying Song
Ziying Song
As a critical task in autonomous driving perception systems, 3D object detection is used to identify and track key objects, such as vehicles and pedestrians. However, detecting distant, small, or occluded objects (hard instances) remains a challenge, which directly compromises the safety of autonomous driving systems. We observe that existing multi-modal 3D object detection methods often follow a single-guided paradigm, failing to account for the differences in information density of hard instances between modalities. In this work, we propose DGFusion, based on the Dual-guided paradigm, which fully inherits the advantages of the Point-guide-Image paradigm and integrates the Image-guide-Point paradigm to address the limitations of the single paradigms. The core of DGFusion, the Difficulty-aware Instance Pair Matcher (DIPM), performs instance-level feature matching based on difficulty to generate easy and hard instance pairs, while the Dual-guided Modules exploit the advantages of both pair types to enable effective multi-modal feature fusion. Experimental results demonstrate that our DGFusion outperforms the baseline methods, with respective improvements of +1.0% mAP, +0.8% NDS, and +1.3% average recall on nuScenes. Extensive experiments demonstrate consistent robustness gains for hard instance detection across ego-distance, size, visibility, and small-scale training scenarios.

8 December 2025 in TechRxiv
Verified
Preprint
Open access
FJFeiyang Jia
Feiyang Jia
CJCaiyan Jia
Caiyan Jia
ZSZiying Song
Ziying Song
ZBZhicheng Bao
Zhicheng Bao
LLLin Liu
Lin Liu
YLYadan Luo
Yadan Luo
Driving World Models (DWM) have garnered significant attention for their capacity to explicitly model vehicle dynamics, integrate multimodal sensor inputs into a unified representation, and facilitate long-horizon reasoning-showing considerable potential to enhance the safety and robustness of autonomous driving systems. This survey offers a comprehensive review of DWM through the central lens of robustness. It begins by outlining the fundamentals and unique role of DWM in autonomous driving, followed by a taxonomy of existing techniques organized by their paradigms, architectures, and downstream applications. A pioneering progressive analysis is then conducted, categorizing the development of DWM robustness into three distinct stages-from Robustness 1.0 to Robustness 3.0. This analysis spans multiple dimensions such as model design, task formulation, evaluation criteria, and adaptation capabilities, while also identifying key opportunities and future research directions toward building robust and generalizable world models. Designed to serve as a structured reference for both researchers and practitioners, this survey synthesizes the current state of the art and highlights persistent challenges in the field.

23 September 2026 in Information Processing & Management
Verified
Journal Article
Closed access
LHLina Han
Lina Han
YQYuhan Qiao
Yuhan Qiao
FJFeiyang Jia
Feiyang Jia
YZYan Zhao
Yan Zhao
CJCaiyan Jia
Caiyan Jia