CVPR2024

HardMo: A Large-Scale Hardcase Dataset for Motion Capture

Jiaqi Liao, Chuanchen Luo, Yinuo Du, Yuxi Wang, Xucheng Yin, Man Zhang, Zhaoxiang Zhang, Junran Peng

摘要

a) (b) (c) (a) (b) (c) Figure 1. Comparisons between vanilla 4DHumans [10] and our HardMo-4DHumans. The teaser comprises four sub-figures. Each subfigure, from left to right, corresponds to (a) the input image, (b) the prediction of HardMo-4DHumans, and (c) the prediction of vanilla 4DHumans. In comparison, our HardMo-4DHumans is superior in the alignment of hand and foot posture.