Vision AI at GTC DC and the AI City Challenge at ICCV 2025
Published:
October brought together our industrial vision AI work and the research community around multi-camera 3D perception. Our Foxconn factory digital-twin demonstration appeared in NVIDIA’s GTC DC 2025 keynote, and I hosted the 9th AI City Challenge workshop at ICCV 2025 in Honolulu.
Highlights
- Factory digital twins: The demonstration combined Metropolis, Cosmos, and Omniverse to analyze activity in a Foxconn factory environment and illustrate spatial monitoring workflows.
- AI City Challenge: The workshop brought together 245 teams from 15 countries, with work spanning multi-camera 3D perception, transportation safety, warehouse analytics, and edge-optimized fisheye detection.
- Benchmark adoption: PAISS and WTS Traffic Safety had nearly 400,000 combined Hugging Face downloads at this workshop checkpoint.
- BMTT keynote: I presented MCBLT/BEV-SUSHI, Sparse4D, and graph-based association for scalable multi-camera tracking at the 8th Workshop on Benchmarking Multi-Target Tracking.
Thank you to the teams across NVIDIA, Yizhou Wang, Sameer Pusegaonkar, and collaborators in Laura Leal-Taixé’s group.

