name: annotation-ui-and-data
description: "Operate AnyLabeling desktop annotation, label JSON, canvas/shape
editing, and annotation export workflows."
disable-model-invocation: true
metadata:
disco-role: operating
license: GPL 3.0
annotation-ui-and-data
Use this sub-skill when the task is about AnyLabeling's manual desktop annotation workflow, label JSON files, image navigation, canvas shape behavior, annotation-related CLI/config flags, or exporting saved labels to YOLO, Pascal VOC, COCO, or CreateML.
Route here for
- Opening images or directories, navigating next/previous images, autosave, output directory/file behavior, and label validation/session options.
- Understanding or repairing AnyLabeling JSON files: top-level fields,
shapes, flags, text, group_id, other_data, imagePath, imageData, and image dimensions.
- Polygon, rectangle, circle, line, point, and linestrip shape editing; grouped shapes; shape text; image-level flags/text; canvas bounded movement and rectangle point ordering.
- Format export behavior, including which shape types are exported or skipped and how dataset-wide exports differ from per-file exports.
Route elsewhere
- Auto-labeling model registry, model downloads, ONNX/CoreML inference, SAM/YOLO prompt behavior, and embedding caches belong to the auto-labeling-models sub-skill.
- Packaging, releases, build resources, translation/resource compilation, PyInstaller, and PyPI wheel variants belong to the packaging-release sub-skill.
- Deep PyQt internals are out of scope unless they affect user-visible annotation behavior or troubleshooting.
Read first
- references/ui-workflows.md for desktop workflow, CLI/config flags, navigation, saving, validation, and canvas behavior.
- references/data-formats-and-export.md for the JSON schema, preservation rules, export APIs, and exporter limitations.
- references/troubleshooting.md for invalid JSON, missing images, dimension mismatches, unsupported export shapes, headless Qt issues, autosave/output confusion, and startup colormap regressions.
Bundled helpers
Both scripts are self-contained, argparse-based, safe by default, and runnable from any current directory.