HumanScore: Benchmarking Human Motions in Generated Videos
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866908986053230592 |
|---|---|
| author | Fang, Yusu Xiang, Tiange Tan, Tian Schuetz, Narayan Delp, Scott Fei-Fei, Li Adeli, Ehsan |
| author_facet | Fang, Yusu Xiang, Tiange Tan, Tian Schuetz, Narayan Delp, Scott Fei-Fei, Li Adeli, Ehsan |
| contents | Recent advances in model architectures, compute, and data scale have driven rapid progress in video generation, producing increasingly realistic content. Yet, no prior method systematically measures how faithfully these systems render human bodies and motion dynamics. In this paper, we present HumanScore, a systematic framework to evaluate the quality of human motions in AI-generated videos. HumanScore defines six interpretable metrics spanning kinematic plausibility, temporal stability, and biomechanical consistency, enabling fine-grained diagnosis beyond visual realism alone. Through carefully designed prompts, we elicit a diverse set of movements at varying intensities and evaluate videos generated by thirteen state-of-the-art models. Our analysis reveals consistent gaps between perceptual plausibility and motion biomechanical fidelity, identifies recurrent failure modes (e.g., temporal jitter, anatomically implausible poses, and motion drift), and produces robust model rankings from quantitative and physically meaningful criteria. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_20157 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | HumanScore: Benchmarking Human Motions in Generated Videos Fang, Yusu Xiang, Tiange Tan, Tian Schuetz, Narayan Delp, Scott Fei-Fei, Li Adeli, Ehsan Computer Vision and Pattern Recognition Recent advances in model architectures, compute, and data scale have driven rapid progress in video generation, producing increasingly realistic content. Yet, no prior method systematically measures how faithfully these systems render human bodies and motion dynamics. In this paper, we present HumanScore, a systematic framework to evaluate the quality of human motions in AI-generated videos. HumanScore defines six interpretable metrics spanning kinematic plausibility, temporal stability, and biomechanical consistency, enabling fine-grained diagnosis beyond visual realism alone. Through carefully designed prompts, we elicit a diverse set of movements at varying intensities and evaluate videos generated by thirteen state-of-the-art models. Our analysis reveals consistent gaps between perceptual plausibility and motion biomechanical fidelity, identifies recurrent failure modes (e.g., temporal jitter, anatomically implausible poses, and motion drift), and produces robust model rankings from quantitative and physically meaningful criteria. |
| title | HumanScore: Benchmarking Human Motions in Generated Videos |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2604.20157 |