Research map / Detection and segmentation / Tracking
LIMMT: Less Is More for Motion Tracking
Yu Guan, Zekun Qi, Chenghuai Lin, Xuchuan Chen, Wenyao Zhang, Jilong Wang, XinQiang Yu, He Wang, Li Yi
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 |
|---|---|---|---|
| train on GQS-curated 3% of AMASS instead of the full dataset | Any2Track · AMASS test | Success Rate: 0.942 -> 0.956 | Improved |
| train on GQS-curated 3% of AMASS instead of the full dataset | TWIST2 · AMASS test | Success Rate: 0.825 -> 0.861 | Improved |
| random 3% subsampling instead of GQS selection (ablation) | Any2Track · AMASS test | Success Rate: 0.942 -> 0.838 | Hurt |
| random 3% subsampling instead of GQS selection (ablation) | TWIST2 · AMASS test | Success Rate: 0.825 -> 0.649 | Hurt |
| add physics filtering at 100% data scale (ablation) | Any2Track · AMASS test | Success Rate: 94.2% -> 95.4% | Improved |
| add physics filter before embedding-based selection at 3% data (ablation) | GQS · AMASS test, 3% data | Success Rate: 0.911 -> 0.956 | Improved |
| add diversity-based FPS to complexity-only selection (ablation) | GQS · AMASS test, 3% data | Success Rate: 0.934 -> 0.956 | Improved |
| add complexity weighting to vanilla FPS (ablation) | GQS · AMASS test, 3% data | Success Rate: 0.946 -> 0.956 | Improved |
| raise data ratio from 10% to 90% of GQS selection (ablation) | Any2Track+GQS · AMASS test | Success Rate | No clear effect |
| train on top physical-score slice only versus mid-score slice (ablation) | Any2Track · AMASS physical score percentile slices | Success Rate: 94.6% (0-10%) vs 96.3% (60-70%) vs 92.2% (90-100%) | Mixed |
| filter out high-velocity or high-jerk motions (ablation) | motion data curation · AMASS sensitivity analysis | tracking performance | Hurt |
| filter out floating and foot-sliding motions (ablation) | motion data curation · AMASS sensitivity analysis | tracking performance: Impact +2.6 floating, +1.0 foot slide | Improved |
Closest papers by meaning
- Universal Humanoid Motion Representations for Physics-Based Control (ICLR 2024 · similarity 0.86)
- Humanoid Horizon: Extending Task Horizon in Whole-Body Loco-Manipulation via Parallel Training, Dynamic Starting, and Reward Gating (NeurIPS 2026 · similarity 0.85)
- PhyMotion: Structured 3D Motion Reward for Physics-Grounded Human Video Generation (NeurIPS 2026 · similarity 0.84)
- SUGAR: A Scalable Human-Video-Driven Generalizable Humanoid Loco-Manipulation Learning Framework (NeurIPS 2026 · similarity 0.84)
- PDF-HR: Pose Distance Fields for Humanoid Robots (NeurIPS 2026 · similarity 0.83)
- DiT4DiT: Jointly Modeling Video Dynamics and Actions for Generalizable Robot Control (NeurIPS 2026 · similarity 0.83)
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.
