annotators-and-preprocessing
Use ControlNet 1.0 annotators and preprocessing safely: Canny, HED, MLSD, MiDaS depth/normal, OpenPose, Uniformer segmentation, HWC3 conversion, resolution rounding, and detector checkpoint expectations.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- annotators-and-preprocessing
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Use ControlNet 1.0 annotators and preprocessing safely: Canny, HED, MLSD, MiDaS depth/normal, OpenPose, Uniformer segmentation, HWC3 conversion, resolution rounding, and detector checkpoint expectations.
- disable-model-invocation
- true
02 资源文件 3
- 参考文档references/annotator-reference.md9.0 KB
- 参考文档references/troubleshooting.md6.9 KB
- 脚本scripts/inspect_annotator_inputs.py8.1 KB
03 SKILL.md 正文
Annotators and Preprocessing
Use this sub-skill when a task asks how to prepare or debug ControlNet conditioning maps, inspect gradio_annotator controls, choose an annotator for a control task, or reason about image channels, resolution multiples, and detector checkpoints.
For generation from an existing conditioning map, route to gradio-inference-apps. For model/checkpoint layout, ControlNet weight files, or config questions, route to model-and-weight-utilities. For Fill50K, paired conditioning datasets, or training preprocessing, route to training-and-datasets.
Read Or Run
- Read references/annotator-reference.md to choose Canny, HED, MLSD, MiDaS depth/normal, OpenPose, or Uniformer and confirm function signatures, parameters, output channels, RGB/BGR cautions, and checkpoint expectations.
- Read references/troubleshooting.md when outputs look inverted, black/blank, wrong-colored, unexpectedly resized, or fail because Gradio, OpenCV, CUDA, optional packages, or detector checkpoints are missing.
- Run scripts/inspect_annotator_inputs.py before writing batch preprocessing code; it checks ControlNet-style
HWC3and 64-multiple resizing on tiny generated arrays or a local input image, and can statically listgradio_annotator.pyfunction signatures when given--repo-root.
Safe Default Approach
- Normalize incoming images to uint8 HWC RGB-style arrays with one, three, or four channels; ControlNet's
HWC3behavior composites RGBA over white and expands grayscale to three channels. - Resize by scaling the short side to the requested resolution, then rounding height and width to multiples of 64; do not assume the result is square unless the aspect ratio already was square.
- Prefer Canny for checkpoint-free, CPU-friendly preprocessing; treat HED, MLSD, MiDaS, OpenPose, and Uniformer as checkpoint- and dependency-dependent detectors.
- Preserve polarity and channel order deliberately: edge maps can be white-on-black or black-on-white depending on downstream expectations, and normal/pose/segmentation visualizations are easy to misinterpret if RGB/BGR is swapped.
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).