data-preparation
Guides Nerfstudio dataset conversion, transforms.json validation, camera conventions, device capture modes, and dataparser selection.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- data-preparation
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Guides Nerfstudio dataset conversion, transforms.json validation, camera conventions, device capture modes, and dataparser selection.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/data-formats.md1.8 KB
- 参考文档references/troubleshooting.md1.8 KB
- 参考文档references/workflows.md2.3 KB
- 脚本scripts/validate_nerfstudio_dataset.py6.6 KB
03 SKILL.md 正文
Data Preparation
Use this route before training when the task is about custom captures, camera
poses, transforms.json, ns-process-data, or choosing a dataparser.
What this route covers
- Converting images, videos, Polycam, Record3D, Metashape, RealityCapture, ODM, and Project Aria style captures into the Nerfstudio dataset layout.
- Validating the Nerfstudio
transforms*.jsonschema, image/depth/mask paths, split lists, and camera conventions. - Distinguishing COLMAP/FFmpeg/hloc/external-tool requirements from package-level Python issues.
- Choosing between
nerfstudio-data,blender-data,minimal-parser,colmap, and other built-in dataparser names.
What this route excludes
- Training method choice and config overrides: use
training-and-configsafter the dataset validates. - Dataset download commands: use
cli-workflowsfirst, because downloads need network and disk checks. - Custom dataparser packaging: use
api-extension.
Read/run these bundled files
references/data-formats.mdfortransforms*.json, coordinate, depth, mask, and split fields.references/workflows.mdfor conversion recipes and dataparser routing.references/troubleshooting.mdfor COLMAP, FFmpeg, path, and schema failures.scripts/validate_nerfstudio_dataset.pyto check a dataset directory ortransforms*.jsonwithout modifying data.
Safe workflow
- Identify the capture type and external prerequisites.
- If using images/video, confirm COLMAP and FFmpeg unless an existing sparse model makes
--skip-colmapvalid. - Run the appropriate
ns-process-datacommand into a new output directory. - Validate the output directory with the bundled validator.
- Pass the dataset directory to
ns-trainthroughtraining-and-configs.
Key reminders
- Nerfstudio uses OpenGL/Blender-style camera coordinates; COLMAP/OpenCV axes differ and are converted by the processing/dataparser code.
- Image paths in
transforms*.jsonare relative to the dataset directory unless absolute paths are intentionally used. - Depth images are specified per frame with
depth_file_pathand are interpreted in millimeters by default. - Masks must be one-channel black/white images, same resolution as RGB, and must be present for all frames if used.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).