< Explain other AI papers

CoMotion: Concurrent Multi-person 3D Motion

Alejandro Newell, Peiyun Hu, Lahav Lipson, Stephan R. Richter, Vladlen Koltun

2025-04-22

CoMotion: Concurrent Multi-person 3D Motion

Summary

This paper talks about CoMotion, a new system that can watch a crowded scene with just one camera and accurately track the 3D movements of several people at the same time.

What's the problem?

The problem is that most systems struggle to follow the movements of many people in busy places, especially when people block each other from view or move around a lot. This makes it hard to keep track of who is who and what each person is doing.

What's the solution?

The researchers created CoMotion, which can detect and follow the 3D positions and movements of multiple people, even when they overlap or get in each other's way. The system is designed to stay accurate and fast, so it works well even in tricky, crowded situations.

Why it matters?

This matters because it can be used for things like sports analysis, security, animation, and virtual reality, where understanding how groups of people move in real time is really important.

Abstract

A system detects and tracks 3D poses of multiple people in crowded scenes using a single camera, maintaining accuracy and speed even with occlusions.