Wenzhao Zheng

MindView: Benchmarking Generative Multi-View Spatial Intelligence

Yanbo Wang, Siqi Luo, Siyu Zhang, et al.

SM4RT: Learning Structured Motion Geometry for 4D Reconstruction

Shing Ho J. Lin, Wenzhao Zheng, Dong Zhuo, et al.

BAMI: Training-Free Bias Mitigation in GUI Grounding

Borui Zhang, Bo Zhang, Bo Wang, et al.

DVGT-2: Vision-Geometry-Action Model for Autonomous Driving at Scale

Sicheng Zuo, Zixun Xie, Wenzhao Zheng, et al.

Vega: Learning to Drive with Natural Language Instructions

Sicheng Zuo, Yuxuan Li, Wenzhao Zheng, et al.

DriveTok: 3D Driving Scene Tokenization for Unified Multi-View Reconstruction and Understanding

Dong Zhuo, Wenzhao Zheng, Sicheng Zuo, et al.

DriveTok is a new tool that converts multi-camera driving scenes into efficient digital 'tokens' (compressed representations) that capture semantic meaning, depth, and 3D spatial information all at once. Unlike existing tokenizers designed for single images, DriveTok is specifically built for autonomous vehicles with multiple cameras, using advanced attention mechanisms to ensure consistency across different camera views. The tokens can be used for various driving-related tasks like reconstructing images, segmenting objects, predicting depth, and understanding 3D space around the vehicle.