Qiao Wu
Algorithm Engineer
Department of Auto Car
Meituan
Beijing, China
Email: qiaowu [dot] joey [at] gmail [dot] com

[Publications]  [Education]  [Experiences]  [Service]  [Talks] 
[GitHub] 

- I am currently working in Beijing as a algorithm engineer at Department of Auto Car, Meituan, focusing on the Autonomous Driving.
- I obtained my M.S. degree at Northwestern Polytechnical University (NWPU), co-supervied by Jiaqi Yang and Mathieu Salzmann. Before learning at NWPU, I received B.Eng. degree at China University of Geosciences, adviced by Kun Sun.
- I am now interested in 3D computer vision, especially focusing on the tasks related to Autonomous Driving. Please send me an email if you have any ideas for cooperation.

NEW We have one paper accepted by CVPR 2025. See you in Nashville!


Publications

Unlocking Generalization Power in LiDAR Point Cloud Registration, CVPR 2025
Zhenxuan Zeng , Qiao Wu>, Xiyu Zhang, Lin Yuanbo Wu, Pei An, Jiaqi Yang, Ji Wang, Peng Wang
paper / bibtex / code


          

SPU-IMR: Self-supervised Arbitrary-scale Point Cloud Upsampling via Iterative Mask-recovery Network, AAAI 2025
Ziming Nie, Qiao Wu, Chenlei Lv, Siwen Quan, Zhaoshuai Qi, Muze Wang, Jiaqi Yang
paper / bibtex / code


          

3D Single-object Tracking in Point Clouds with High Temporal Variation, ECCV 2024
Qiao Wu, Kun Sun, Pei An, Mathieu Salzmann, Yanning Zhang, Jiaqi Yang

We explore a new task in 3D SOT, and presented the first 3D SOT framework for high temporal variation scenarios, HVTrack. Its three main components, RPM, BEA, and CPA, allow HVTrack to achieve robustness to point cloud variations, similar object distractions, and background noise. HVTrack significantly outperforms existing trackers in high temporal variation scenarios (11.3% and 15.7% improvement in success and precision at medium intensity of variation). The performance gap between our HVTrack and existing trackers widens as variations are exacerbated. It also surpasses existing methods in both nuScenes and Waymo benchmarks of regular tracking, achieving SOTA.

paper / bibtex / code


          

MixCycle: Mixup Assisted Semi-Supervised 3D Single Object Tracking with Cycle Consistency, ICCV 2023
Qiao Wu, Jiaqi Yang, Kun Sun, Chu'ai Zhang, Yanning Zhang, Mathieu Salzmann

We propose the first semi-supervised approach to 3D Single Object Tracking. Our method, MixCycle, uses self and forward-backward cycle-consistency for supervision, and introduce a data augmentation strategy named SOTMixup to improve the tracker's robustness to point cloud diversity. It is generalizes to appearance matching-based trackers.

paper / bibtex / code


          

MAC: Maximal Cliques for 3D Registration, TPAMI 2024
Jiaqi Yang, Xiyu Zhang, Peng Wang, Yulan Guo, Kun Sun, Qiao Wu, Shikun Zhang, Yanning Zhang

This paper presents a 3D registration method with maximal cliques (MAC) for 3D point cloud registration (PCR). The key insight is to loosen the previous maximum clique constraint and mine more local consensus information in a graph for accurate pose hypotheses generation. In addition, we present a variant of MAC if given overlap prior, called MAC-OP. Extensive experiments demonstrate that both MAC and MAC-OP effectively increase registration recall, outperform various state-of-the-art methods, and boost the performance of deep-learned methods.

paper / bibtex / code


          


Education


Experiences


Professional service

  • Program reviewing:
  • Talks


    Award