Abstract
Despite rapid progress in video generation models, they still exhibit obvious motion deficiencies, often manifested as incorrect object motion. However, most existing video quality evaluations focus on aesthetic quality or text-video alignment. To address this gap, we study object-centric motion fidelity assessment, evaluating target objects along object consistency, motion continuity, and physical plausibility. To achieve this, we first introduce VidMotion, a diagnostic dataset of 6,879 videos with designated moving objects and fine-grained annotations including dimension-wise scores and failure causes. We further propose MotionInsight, a diagnostic evaluator that shifts assessment from implicit RGB-frame observation to explicit motion-space diagnosis. By constructing motion-aware representations, MotionInsight makes subtle motion deficiencies more observable. We also introduce motion-specific rewards during GRPO to transform observed motion into a diagnostic assessment. Experiments demonstrate that MotionInsight provides an effective basis for diagnosing object motion deficiencies, producing human-aligned scores along three dimensions and grounded explanations.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Zhan, J., Ma, Y., Xing, Q., Zhang, X., Tong, J., Li, J., Zhang, L., Zhao, S., & Xue, T. (2026). MotionInsight: Diagnosing Object Motion Deficiencies in Generated Videos. https://omanscience.com/en/articles/motioninsight-diagnosing-object-motion-deficiencies-in-generated-videos
MLA 9
Zhan, Jiahao, et al. "MotionInsight: Diagnosing Object Motion Deficiencies in Generated Videos." https://omanscience.com/en/articles/motioninsight-diagnosing-object-motion-deficiencies-in-generated-videos.
Chicago (author–date)
Zhan, Jiahao, Yongrui Ma, Qunliang Xing, Xuanyu Zhang, Jingqi Tong, Junlin Li, Li Zhang, Shijie Zhao, and Tianfan Xue. 2026. "MotionInsight: Diagnosing Object Motion Deficiencies in Generated Videos." https://omanscience.com/en/articles/motioninsight-diagnosing-object-motion-deficiencies-in-generated-videos.
Harvard
Zhan, J., Ma, Y., Xing, Q., Zhang, X., Tong, J., Li, J., Zhang, L., Zhao, S. and Xue, T. (2026) 'MotionInsight: Diagnosing Object Motion Deficiencies in Generated Videos', Available at: https://omanscience.com/en/articles/motioninsight-diagnosing-object-motion-deficiencies-in-generated-videos.
Vancouver
Zhan J, Ma Y, Xing Q, Zhang X, Tong J, Li J, et al. MotionInsight: Diagnosing Object Motion Deficiencies in Generated Videos. https://omanscience.com/en/articles/motioninsight-diagnosing-object-motion-deficiencies-in-generated-videos
IEEE
J. Zhan, Y. Ma, Q. Xing, X. Zhang, J. Tong, J. Li, L. Zhang, S. Zhao, and T. Xue, "MotionInsight: Diagnosing Object Motion Deficiencies in Generated Videos," https://omanscience.com/en/articles/motioninsight-diagnosing-object-motion-deficiencies-in-generated-videos.