conversion
Convert common annotation and detector-output formats into mAP evaluator text files.
- License 许可
- Apache-2.0
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- conversion
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Convert common annotation and detector-output formats into mAP evaluator text files.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档references/data-formats.md4.8 KB
- 参考文档references/troubleshooting.md5.9 KB
- 参考文档references/workflows.md6.5 KB
- 脚本scripts/convert_annotations.py25.5 KB
03 SKILL.md 正文
Conversion
Use this sub-skill when the user needs annotation or detector-output files transformed into the evaluator's required per-image text files before running mAP/AP evaluation.
What this sub-skill covers
- Ground-truth conversions:
- PASCAL VOC XML object annotations to
ground-truth/*.txtrows. - YOLO normalized label files to
ground-truth/*.txtrows using a class list and image dimensions. - keras-yolo3 ground-truth annotation files to one evaluator text file per image.
- PASCAL VOC XML object annotations to
- Detection-result conversions:
- darkflow JSON detections to
detection-results/*.txtrows. - darknet
result.txtdetector output todetection-results/*.txtrows. - keras-yolo3 detection annotation files to one evaluator text file per image.
- darkflow JSON detections to
- Safer conversion behavior than the legacy one-off scripts: explicit input paths, explicit output directories, no moving or renaming source files.
Boundaries and routing
- Do not compute AP/mAP metrics here. After conversion, route evaluator execution to the sibling
evaluationsub-skill. - Do not repair mismatched ground-truth/detection-result file sets here. Route class lookup, intersection checks, and repair workflows to the sibling
data-validationsub-skill. - Do not depend on the original repository checkout. Use the bundled helper and bundled references in this sub-skill.
Start here
- Read references/data-formats.md to identify the source format and the exact evaluator row format you must produce.
- Follow references/workflows.md for the matching conversion command.
- If the conversion fails, use references/troubleshooting.md before changing data or assumptions.
- Use the bundled converter: scripts/convert_annotations.py.
Quick route table
| User request | Route |
|---|---|
| "Convert VOC XML annotations" | voc-xml-gt workflow in references/workflows.md. |
| "Convert YOLO labels for ground truth" | yolo-gt workflow; require class list plus image dimensions. |
| "Convert darkflow JSON detections" | darkflow-json-dr workflow. |
| "Convert darknet result.txt" | darknet-result-dr workflow. |
| "Convert keras-yolo3 annotations" | keras-yolo3 --gt or keras-yolo3 --dr; choose flat or recursive output layout. |
Output contract
Successful conversion produces evaluator-ready .txt files in an explicit output directory. Each output file represents one image. Ground-truth and detection-result files should later have matching basenames before evaluation; checking and repairing that intersection belongs to data-validation.
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).