Views
No views yet
1from lib.model.load_critic import load_critic
2from parsedata import into_critic
3import torch
4
5device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
6critic_model = load_critic("motioncritic_pre.pth", device)
7
8# motion: [bs, 25, 6, frame], rot6d
9preprocessed = into_critic(motion) # [bs, frame, 25, 3], axis-angle
10scores = critic_model.module.batch_critic(preprocessed)1@inproceedings{motioncritic2025,
2 title={Aligning Motion Generation with Human Perceptions},
3 author={Wang, Haoru and Zhu, Wentao and Miao, Luyi and Xu, Yishu and Gao, Feng and Tian, Qi and Wang, Yizhou},
4 booktitle={International Conference on Learning Representations (ICLR)},
5 year={2025}
6}