MITracker: Multi-View Integration for Visual Object Tracking

Mengjie Xu*, Yitao Zhu*, Haotian Jiang, Jiaming Li, Zhenrong Shen, Sheng Wang, Haolin Huang, Xinyu Wang, Han Zhang, Qing Yang, Qian Wang⁺

• Introduces MVTrack, a large-scale dataset with 234K frames and precise annotations for 27 object categories, providing a benchmark for class-agnostic multi-view object tracking.
• Proposes MITracker, a method leveraging BEV-guided 3D feature volumes and spatial-enhanced attention for robust target recovery in multi-view tracking.
• Demonstrates that MITracker achieves state-of-the-art performance, improving recovery rates from 56.7% to 79.2% on MVTrack dataset.

MUC: Mixture of Uncalibrated Cameras for Robust 3D Human Body Reconstruction

Yitao Zhu*, Sheng Wang*, Mengjie Xu, Zixu Zhuang, Zhixin Wang, Kaidong Wang, Han Zhang, Qian Wang⁺

• Introduces a technique for accurately reconstructing 3D human poses and shapes from images captured by uncalibrated cameras.
• Utilizes pre-trained monocular models to estimate camera positions and employs a distance ranking optimization strategy for precise joint fusion, addressing self-occlusion issues.
• Deploy a model to reweight human surfaces for accurate body shape estimation outputs.