菜单
Audit YOLO dataset quality — class distribution, annotation quality, image stats, and improvement suggestions.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | review-dataset |
| description | Audit YOLO dataset quality — class distribution, annotation quality, image stats, and improvement suggestions. |
Dataset quality audit + profiling for architecture selection.
Read yolo-project.yaml for dataset path, class names, and imgsz.
If no config, ask user for dataset path.
yolo-validate <dataset_path> --strict
Scan label files for:
Sample random images and visually verify annotation quality using Claude's multimodal vision.
python scripts/draw_annotations.pyInclude visual spot-check results in the audit report with:
Run the profiling script with imgsz from yolo-project.yaml (default 640):
python scripts/profile_dataset.py \
--labels <dataset>/labels/train \
--images <dataset>/images/train \
--imgsz <imgsz> \
--class-names "<comma-separated from data.yaml>"
The script outputs structured YAML with:
Save the output to experiments/dataset_profile.yaml.
Fill the Dataset Summary section in training-plan.md with profile data:
Create experiments/dataset_audit.md with:
Key findings, profile highlights, and architecture suggestion.
Set up autonomous training monitoring — creates cron jobs to track long-running training, auto-continue pipeline when training completes.
Orchestrate the full active learning loop: train, analyze, push to CVAT, wait for review, pull, merge, retrain.
Analyze YOLO training runs — compares to baseline/best, checks per-class regression, analyzes training dynamics and tune convergence, writes actionable recommendations.
Run autonomous YOLO training experiments — reads training-plan.md, assesses bottlenecks, acts strategically, and delegates HP optimization to model.tune().
Profile YOLO model inference speed, FPS, and size across image sizes and export formats.
Initialize a new YOLO project — detects your dataset's starting state and routes through the right tools.