Physical AI Smart Spaces: A Large-Scale Benchmark for Multi-Camera 3D Perception in Smart Spaces
Published in 2026 Conference on Neural Information Processing Systems, 2026
Recommended citation: Yuxing Wang, Yizhou Wang, Anqi Li, Shuo Wang, Sameer Satish Pusegaonkar, Haoquan Liang, Jiajun Li, Shenxin Jiang, Jianhe Yuan, Shangru Li, Tongwei Dai, Zihao Chen, David C. Anastasiu, Sujit Biswas, Xunlei Wu and Zheng Tang. "Physical AI Smart Spaces: A Large-Scale Benchmark for Multi-Camera 3D Perception in Smart Spaces". Proceedings of Conference on Neural Information Processing Systems (NeurIPS 2026). 2026.
[Dataset] [Conference]
Abstract
Physical AI Smart Spaces (PAISS) is a large-scale benchmark for multi-class, multi-camera 3D perception in indoor smart spaces. It contains over 280 hours of synchronized 1080p footage from nearly 1,800 cameras across warehouses, hospitals, retail venues, and related environments, together with multi-camera identities, 2D and 3D bounding boxes, camera calibration, and depth where available. The benchmark spans Isaac Sim synthetic generation, Cosmos Transfer appearance augmentation, and real-world Sim2Real evaluation. It also provides standardized benchmark protocols, an official evaluation system, and a 3D extension of Higher Order Tracking Accuracy for evaluating 3D locations and boxes.
BibTeX
@InProceedings{Wang26PAISS,
author = {Yuxing Wang and Yizhou Wang and Anqi Li and Shuo Wang and Sameer Satish Pusegaonkar and Haoquan Liang and Jiajun Li and Shenxin Jiang and Jianhe Yuan and Shangru Li and Tongwei Dai and Zihao Chen and David C. Anastasiu and Sujit Biswas and Xunlei Wu and Zheng Tang},
title = {Physical {AI} {S}mart {S}paces: {A} large-scale benchmark for multi-camera {3D} perception in smart spaces},
booktitle = {Proc. NeurIPS},
address = {Atlanta, GA, USA},
year = {2026}
}