retargeting-and-motion-data
Prepare ProtoMotions MotionLib data, conversion pipelines, PyRoki retargeting, contact labels, FPS handling, and motion utility checks.
- License 许可
- Apache-2.0
- 最近更新
- 2026-10-08
- Stars
- ★ 11
- Forks
- 3
01 前言 Front Matter
- name
- retargeting-and-motion-data
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Prepare ProtoMotions MotionLib data, conversion pipelines, PyRoki retargeting, contact labels, FPS handling, and motion utility checks.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/conversion-utilities.md2.1 KB
- 参考文档references/motion-data-formats.md2.0 KB
- 参考文档references/retargeting-pipeline.md2.6 KB
- 参考文档references/troubleshooting.md1.8 KB
03 SKILL.md 正文
ProtoMotions retargeting and motion data
Use this sub-skill for MotionLib schemas, AMASS/PHUMA/SEED/Kimodo data preparation, SMPL/SOMA/G1/H1 retargeting, PyRoki pipeline reasoning, contact-label handling, FPS/downsampling, motion filtering, and packaged-motion utilities.
Read first
references/motion-data-formats.md:.motion, MotionLib.pt, YAML lists, contacts, FPS, and body alignment.references/retargeting-pipeline.md: SMPL/SOMA keypoint extraction, PyRoki retargeting, contacts, conversion, and packaging sequence.references/conversion-utilities.md: useful conversion/helper surfaces and when to run or skip them.references/troubleshooting.md: data, contact, filter, and environment failures.scripts/subset_motion_lib.py: CPU utility to sample every Nth motion from a packaged MotionLib.scripts/summarize_motion_lib.py: safe MotionLib.ptstructure summary.
Decision flow
- Identify source skeleton/data: AMASS SMPL, SMPL-X, SOMA, PHUMA, SEED G1 CSV, Kimodo-generated CSV/NPZ, or already packaged MotionLib.
- Confirm target robot/skeleton:
smpl,soma23,g1, orh1_2are the main evidence-backed routes. - For robot retargeting, use separate ProtoMotions and PyRoki environments.
- Preserve source motion contacts when retargeting whenever the pipeline provides them; re-computing contacts from imperfect retargeted output can be less reliable.
- Package final
.motionfiles into a MotionLib.ptbefore training. - Start with a small subset (
skip_freqorsubset_motion_lib.py) before full AMASS/SEED scale.
Safe utilities
python scripts/summarize_motion_lib.py <motion_lib.pt>
python scripts/subset_motion_lib.py <input.pt> <output.pt> --sample-every 200
These helpers operate on local .pt files and do not require simulator execution.
Retargeting cautions
- PyRoki optimizes trajectories and may require CUDA/JAX dependencies separate from ProtoMotions.
- The full AMASS-to-robot pipeline can be long and data-heavy; use skip/subset parameters for smoke tests.
- Contact labels must match motion length after FPS downsampling.
- Motion filters can drop invalid outputs; inspect thresholds before assuming conversion failed.
- If working from a package-only install, require user-provided conversion scripts or a source checkout because some data-prep scripts are not part of the installed package.
04 同仓库更多 Skills
obsidian-clipper-template-creator
Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.
obsidian-clipper-template-creator
Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.
python-api
Programmatic Object-Detection-Metrics API guidance for BoundingBox, BoundingBoxes, Evaluator, enums, coordinate conversion, AP/mAP semantics, and source-style import diagnostics.
file-evaluation
Folder-based PASCAL VOC AP/mAP evaluation from ground-truth and detection text files.
object-detection-metrics
Guides legacy Object-Detection-Metrics PASCAL VOC AP/mAP evaluation from detection text folders and direct Python API classes.
nx-workspace-patterns
Configure and optimize Nx monorepo workspaces. Use when setting up Nx, configuring project boundaries, optimizing build caching, or implementing affected commands.