Spatial AI Lab · ShanghaiTech University

Yujiao Shi

Assistant Professor

3D vision across altitudes.
World models from aerial perspectives.

My research connects satellite, aerial, and ground observations to understand and model large-scale environments in 3D. I work on cross-view localization, 3D reconstruction, panoramic vision, and scene generation.

Email ↗Google Scholar ↗CV ↗GitHub ↗

Research vision

One world. Many perspectives.

Connecting views across heights and scales to build a coherent understanding of the world.

Research vision: satellite, aerial, and ground observations connect through cross-view geometry to 3D scene representations and novel-view generation.
Space, air, and ground: complementary observations of a shared 3D world.
01 / GEOMETRY

3D vision across altitudes

Connecting satellite, drone, and ground imagery for camera localization and joint 3D reconstruction, with an emphasis on geometric consistency and learning with limited supervision.

02 / REPRESENTATION

Panoramic vision

Learning geometry and scene representations from 360° imagery for pose estimation, feed-forward reconstruction, and novel view synthesis.

03 / GENERATION

Cross-view scene generation

Bringing geometry and generative modeling together to synthesize consistent ground-level views from satellite imagery and model scenes across viewpoints.

My earlier research established a focus on ground-to-satellite camera localization, progressing from city-scale retrieval to fine-grained pose estimation, and from fully supervised learning to weak supervision. Alongside localization, I have explored satellite-to-ground image synthesis, linking geometric reasoning with visual generation.

Featured research

Geometry, reconstruction & generation

All publications
View all publications

Spatial AI Lab · Opportunities

Explore the world
with us.

I welcome motivated students and researchers who want to connect geometry, learning, and generation to understand the world across viewpoints.

  • Aerial & cross-view 3D vision
  • Panoramic reconstruction
  • Generative scene modeling
  • World models from aerial perspectives

Undergraduate and master's students, research assistants, visiting students, and postdoctoral researchers are welcome to reach out with a CV and a brief description of their research interests.

Get in touch

From the lab

News

  • Area Chair service: NeurIPS 2024–2026, CVPR 2026, ICLR 2027, and 3DV 2027.
  • Cross-View Yaw Estimation at ECCV 2026 (Award Candidate). Congrats, Taeho!
  • Cross3R at NeurIPS 2026. Congrats, Qiwei and Zhongyao!
  • VGA, PanoVGGT, and AffordGrasp at CVPR 2026. Congrats, Tao Jun, Yijing, and Xiaofei!
  • CylinderSplat and SatDreamer360 at ICLR 2026. Congrats, Qiwei and Xianghui!
  • BevSplat at NeurIPS 2025 (Spotlight); GeoDistill at ICCV 2025; CityGo at SIGGRAPH Asia 2025. Congrats, Qiwei, Shaowen, and Weihang!
Earlier news
  • 04/2025: Our ICCV 2025 workshop: From Street to Space: 3D Vision Across Altitudes.
  • 01/2025: One paper accepted by ICLR 2025. Congrats, Xianghui!
  • 11/2024: Two papers accepted by 3DV 2025.
  • 07/2024: One paper accepted by SIGGRAPH Asia 2024 and two papers by ECCV 2024.
  • 04/2024: Our workshop proposal “UAVs in Multimedia” was accepted by ACM MM 2024.
  • 02/2024: Co-organizing the Women in Computer Vision Workshop at ECCV 2024.
  • 01/2024: One paper accepted by ICRA 2024.
  • 10/2023: Co-organizer of “UAVs in Multimedia” at ACM MM 2023.
  • 09/2023: One paper accepted by NeurIPS 2023.
  • 06/2023: One paper accepted by IROS 2023; tutorial speaker on cross-modal camera localization at CVPR 2023.

Community

Academic service

Area Chair
NeurIPS 2024–2026 · CVPR 2026 · ICLR 2027 · 3DV 2027
Workshops
From Street to Space: 3D Vision Across Altitudes, ICCV 2025.
UAVs in Multimedia, ACM MM 2023–2024.
Women in Computer Vision, ECCV 2024.
Recognition
Outstanding Reviewer: CVPR 2021, NeurIPS 2023, CVPR 2024.