Research map / Video / Video generation
Cross-Axis Feature Fusion with Joint-Wise Motion Difference Prediction for Text-Based 3D Human Motion Editing
Gyojin Han, Junmo Kim
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 Soft-DTW joint delta task to motion similarity task (ablation) | Ours · MotionFix (Batch, generated-to-target) | R@1: 71.04 -> 74.38 | Improved |
| add L2 joint delta task to motion similarity task (ablation) | Ours · MotionFix (Batch, generated-to-target) | R@1: 71.04 -> 71.04 | No clear effect |
| use Soft-DTW instead of L2 for joint delta (ablation) | Ours · MotionFix (Batch, generated-to-target) | R@1: 71.04 -> 74.38 | Improved |
| add Soft-DTW joint delta task without motion similarity (ablation) | Ours · MotionFix (Test Set, generated-to-target) | R@1: 30.24 -> 28.85 | Hurt |
| add motion similarity task to Soft-DTW joint delta (ablation) | Ours · MotionFix (Batch, generated-to-target) | R@1: 72.92 -> 74.38 | Improved |
| set Soft-DTW gamma 10 instead of 50 (ablation) | Ours · MotionFix (Batch, generated-to-target) | R@1: 74.79 -> 74.38 | Mixed |
| add explicit time-anchored text condition (ablation) | Ours · MotionFix (Batch, generated-to-target) | R@1: 74.38 -> 70.62 | Hurt |
| add explicit joint-anchored text condition (ablation) | Ours · MotionFix (Batch, generated-to-target) | R@1: 74.38 -> 72.08 | Hurt |
| add explicit axis-fusion condition (ablation) | Ours · MotionFix (Batch, generated-to-target) | R@1: 74.38 -> 70.00 | Hurt |
| use single cross-attention fusion instead of larger fusion block (ablation) | Ours · MotionFix (Batch, generated-to-target) | R@1: 72.92 -> 74.38 | Improved |
| apply motion similarity head after fusion instead of before (ablation) | Ours · MotionFix (Batch, generated-to-target) | R@1: 70.00 -> 74.38 | Improved |
| use Soft-DTW instead of classic DTW (ablation) | Ours · MotionFix (Batch, generated-to-target) | R@1: 72.71 -> 74.38 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SimMotionEdit — SimMotionEdit: Text-Based Human Motion Editing with Motion Similarity Prediction (CVPR 2025) · better on 8 of 8 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SimMotionEdit: Text-Based Human Motion Editing with Motion Similarity Prediction (CVPR 2025 · similarity 0.93)
- InterEdit: Navigating Text-Guided Multi-Human 3D Motion Editing (ECCV 2026 · similarity 0.90)
- SALAD: Skeleton-aware Latent Diffusion for Text-driven Motion Generation and Editing (CVPR 2025 · similarity 0.90)
- Dynamic Motion Blending for Versatile Motion Editing (CVPR 2025 · similarity 0.90)
- MMM: Generative Masked Motion Model (CVPR 2024 · similarity 0.88)
- MotionMix: Weakly-Supervised Diffusion for Controllable Motion Generation (AAAI 2024 · similarity 0.86)
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.
