Qi Zhang 张琦 iMUSE Lab · SZU
Qi Zhang

I lead the iMUSE Lab at Shenzhen University, which I founded in 2022. We study multi-view understanding of large, dynamic scenes: crowd analysis and simulation, robust 3D scene understanding, and emerging work on camera-only 4D occupancy forecasting. Our recent work has moved from measuring where people are to tracking, reconstructing and modelling how those scenes evolve.

Crowd Analysis and Simulation3D Reconstruction and GenerationAutonomous DrivingLarge-scene Point Cloud Analysis
13Graduate researchers
7Alumni
3 + 1PI grants + national R&D sub-project
21Peer-reviewed papers

Recent group work has appeared at CVPR, ICLR, IJCV, ECCV and AAAI, with students as first authors or key contributors across the lab’s directions. Cited by 561 (Google Scholar, 2026-09-27).

Research

Full agenda →

Multi-view crowd intelligence

Counting, detection, localization, tracking and simulation of people seen by many cameras at once — fused onto one ground plane, without per-scene calibration, across views and scenes, and with as little labeled data as possible.

Core programme · 16 papers

Robust 3D scene understanding

Recovering reliable geometry when the observations are uncooperative: badly placed or transformed views, a single image, heavy occlusion across cameras, and city-scale scenes too large to process uniformly.

Core programme · 5 papers

Emerging: street scenes and 4D occupancy

Camera-only forecasting of how occupied space evolves, and cross-view consistent generation of the street scenes that train and test such models.

2 papers · preprints, in progress

Other applications & collaborations

The same fusion machinery pointed at conservation and agriculture — dolphins in the open sea, poultry in a field — and the hyperspectral work this line began from.

4 papers

Selected Publications

All 27 papers →

Recent work from the group; peer-reviewed only, preprints appear on the publications page. Bold marks my name; * denotes corresponding author.

Figure from Multi-view Crowd Tracking Transformer with View-Ground Interactions Under Large Real-World Scenes
Qi Zhang, Jixuan Chen, Kaiyi Zhang, Xinquan Yu, Antoni B. Chan and Hui Huang
CVPR 2026pp. 13626-13635PDFarXivCode
Figure from SynMVCrowd: A Large Synthetic Benchmark for Multi-view Crowd Counting and Localization
Qi Zhang, Daijie Chen, Yunfei Gong and Hui Huang*
IJCV 2026134(4): 1912 citationsarXivPublisherCode
Figure from Mahalanobis Distance-based Multi-view Optimal Transport for Multi-view Crowd Localization
Qi Zhang, Kaiyi Zhang, Antoni B. Chan and Hui Huang*
ECCV 202415 citationsPDFarXivProject

Where this programme started

Group & Openings

iMUSE Lab →

iMUSE Lab takes Ph.D. and master’s students every year and hosts visiting undergraduates. Current openings are in crowd simulation and generation, multi-view 3D reconstruction, and 4D occupancy forecasting. The 2026 Ph.D. intake is full; enquiries for 2027 are welcome now, and pre-admission follows the college’s own schedule.

Students & alumni What we look for Apply / inquire

Academic Service

Full list →

Program committee: CVPR, AAAI, ICCV, NeurIPS, ECCV, ICLR, ICML, IJCAI, ACM MM, SIGGRAPH, EUROGRAPHICS, PG, CVM, CGI, WACV, ACCV, ICPR, ICME, 3DV, ACM MMAsia. Regular journal reviewer for IJCV, TIP, TNNLS, TMM, TCSVT, Information Fusion and others. Committee member: Technical Committee on 3D Vision, China Society of Image and Graphics (CSIG); Technical Committee on Multimodal Interaction and Cognitive Simulation, China Simulation Society.