Qitai Wang

I am currently a fifth-year Ph.D. student at the NLPR, Institute of Automation, Chinese Academy of Sciences (CASIA), supervised by Prof. Tieniu Tan and Prof. Zhaoxiang Zhang. Prior to that, I obtained my Bachelor's degree in Automation from the Department of Automation, Tsinghua University in 2020. Additionally, I have interned at TuSimple.

My research interests involves computer vision, 3D perceptions, video generation models and driving simulation. I am currently exploring fully generative driving simulation.

Email / Google Scholar / Github / Curriculum Vitae

News

2024-09: One paper on 3D perception is accepted to TPAMI.

2025-01: One paper on scene reconstruction-based driving simulation is accepted to CVPR 2025..

2025-01: One paper on generative driving simulation is accepted to ICLR 2025.

2024-09: One paper on driving world model is accepted to NeurIPS 2024 Dataset Track.

2024-07: One paper on End-to-End Multi-Object Tracking is accepted to ECCV 2024.

Research

* indicates equal contribution

	FreeSim: Toward Free-viewpoint Camera Simulation in Driving Scenes Lue Fan, Hao Zhang, Qitai Wang, Hongsheng Li, Zhaoxiang Zhang CVPR 2025 [paper] [Page] After FreeVS we propose FreeSim, a generation-reconstruction hybrid method for free-viewpoint camera simulation, taking the best of two worlds!
	FreeVS: Generative View Synthesis on Free Driving Trajectory Qitai Wang, Lue Fan, Yuqi Wang, Yuntao Chen, Zhaoxiang Zhang ICLR, 2025 [paper] [Page] [code] FreeVS is the first method that supports high-quality generative view synthesis on free driving trajectory. A crucial step towards achieving generative driving simulation.
	DrivingDojo Dataset: Advancing Interactive and Knowledge-Enriched Driving World Model Yuqi Wang, Ke Cheng, Jiawei He, Qitai Wang**, Hengchen Dai, Yuntao Chen, Fei Xia, Zhaoxiang Zhang NeurIPS, 2024, D&B Track [paper] [Page] [code] DrivingDojo dataset features video clips with a complete set of driving maneuvers, diverse multi-agent interplay, and rich open-world driving knowledge, laying a stepping stone for future world model development.
	OneTrack: Demystifying the Conflict Between Detection and Tracking in End-to-End 3D Trackers Qitai Wang, Jiawei He, Yuntao Chen, Zhaoxiang Zhang ECCV, 2024 [paper] We have completely resolved the challenge where the perception performance of end-to-end multi-object tracking was inferior to that of standalone detectors, enabling lossless unification of detection and tracking tasks.
	Uncertain Object Representation for Image-Based 3D Object Perception Qitai Wang, Yuntao Chen, Zhaoxiang Zhang TPAMI [paper] We propose the uncertain representation of 3D objects to meet the indeterminacy of localizing objects in images.
	Immortal Tracker: Tracklet Never Dies Qitai Wang, Yuntao Chen, Ziqi Pang, Naiyan Wang, Zhaoxiang Zhang arxiv [paper] [code] Still the Best, Fastest, Simplest LiDAR-based 3D multi-object tracker so far.

Website Template