Research map / Video / Video generation
MotionHiFlow: Text-to-Motion via Hierarchical Flow Matching
Heng Li, Xiaotong Lin, Ling-An Zeng, Yulei Kang, Shuai Li, Jian-Fang Hu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add hierarchical scales [1/3,2/3,1] instead of single scale [1] (ablation) | MotionHiFlow · HumanML3D (scale ablation) | FID: 0.051 -> 0.032 | Improved |
| add hierarchical scales [1/3,2/3,1] instead of single scale [1] (ablation) | MotionHiFlow · HumanML3D (scale ablation) | MultiModal Dist: 2.723 -> 2.691 | Improved |
| use two scales [1/2,1] instead of single scale [1] (ablation) | MotionHiFlow · HumanML3D (scale ablation) | FID: 0.051 -> 0.038 | Improved |
| use three scales instead of two scales (ablation) | MotionHiFlow · HumanML3D (scale ablation) | FID: 0.038 -> 0.032 | Improved |
| use four scales [1/4,2/4,3/4,1] instead of three (ablation) | MotionHiFlow · HumanML3D (scale ablation) | FID: 0.032 -> 0.035 | No clear effect |
| train at single coarse scale 0.4 instead of full scale (ablation) | MotionHiFlow · HumanML3D (scale ablation) | MultiModal Dist: 2.723 -> 2.717 | No clear effect |
| train at single coarse scale 0.4 instead of full scale (ablation) | MotionHiFlow · HumanML3D (scale ablation) | FID: 0.051 -> 0.106 | Hurt |
| add TMDiT to the transformer baseline (ablation) | Baseline · HumanML3D (component ablation) | FID: 0.074 -> 0.045 | Improved |
| add TMDiT to the transformer baseline (ablation) | Baseline · HumanML3D (component ablation) | MultiModal Dist: 3.043 -> 2.738 | Improved |
| add TMDiT to the transformer baseline (ablation) | Baseline · HumanML3D (component ablation) | R@1: 0.511 -> 0.557 | Improved |
| add topology-aware VAE on top of TMDiT (ablation) | MotionHiFlow · HumanML3D (component ablation) | FID: 0.045 -> 0.032 | Improved |
| add topology-aware VAE on top of TMDiT (ablation) | MotionHiFlow · HumanML3D (component ablation) | MultiModal Dist: 2.738 -> 2.691 | Improved |
Closest papers by meaning
- Towards Detailed Text-to-Motion Synthesis via Basic-to-Advanced Hierarchical Diffusion Model (AAAI 2024 · similarity 0.88)
- Seamless Human Motion Composition with Blended Positional Encodings (CVPR 2024 · similarity 0.88)
- SALAD: Skeleton-aware Latent Diffusion for Text-driven Motion Generation and Editing (CVPR 2025 · similarity 0.88)
- AMD: Anatomical Motion Diffusion with Interpretable Motion Decomposition and Fusion (AAAI 2024 · similarity 0.88)
- SimMotionEdit: Text-Based Human Motion Editing with Motion Similarity Prediction (CVPR 2025 · similarity 0.88)
- MMM: Generative Masked Motion Model (CVPR 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
