CVPR2023

NeMo: 3D Neural Motion Fields from Multiple Video Instances of the Same Action

Kuan-Chieh Wang, Zhenzhen Weng, Maria Xenochristou, João Pedro Araújo, Jeffrey Gu, C. Karen Liu, Serena Yeung

Abstract

Figure 1. We propose the Neural Motion (NeMo) field that learns the global 3D motion from multiple video instances of the same action. NeMo can accurately recover the dynamic ranges of athletic motion. To illustrate, VIBE, a baseline video-based HMR method, fails to capture the large step taken by the subject in the "Baseball Pitch" example, and swaps the arms in "Tennis Serve".