← 返回 Skills 市场

Image Annotation QC

Name: Image Annotation QC
Author: mingo-318

作者 Mingo_318 · GitHub ↗ · v1.0.7

cross-platform ✓ 安全检测通过

353

总下载

当前安装

版本数

在 OpenClaw 中安装

/install image-annotation-qc

功能描述

Image Annotation Quality Control Tool - Automatically detect quality issues in bounding box and polygon segmentation annotations, generate visual reports. Su...

使用说明 (SKILL.md)

Image Annotation QC Tool

Automatically detect quality issues in image annotations and generate detailed reports with visualization.

Features

✅ Multi-format Support: COCO / YOLO / VOC / LabelMe
✅ Auto-detection: Automatically identify format without specification
✅ Scene-specific QC: general / road / industrial / security
✅ Auto-save Reports: Save to qc_report/ directory automatically
✅ Visualization: Generate images with error annotations (red boxes)
✅ Multi-format Reports: TXT / JSON / Excel

Quick Start

# Basic usage (auto-detect format)
python3 scripts/qc_tool.py -i \x3Cimage_dir> -a \x3Cannotation_dir>

# Specify format
python3 scripts/qc_tool.py -i ./images -a ./annotations -f labelme

# Industrial inspection scenario
python3 scripts/qc_tool.py -i ./images -a ./annotations -d industrial

# Sample 100 images
python3 scripts/qc_tool.py -i ./images -a ./annotations -s 100

# Specify output directory
python3 scripts/qc_tool.py -i ./images -a ./annotations -o ./my_report

# Generate Excel report
python3 scripts/qc_tool.py -i ./images -a ./annotations --formats txt json xlsx

Parameters

Parameter	Short	Description	Default
`--image-dir`	`-i`	Image directory	(required)
`--annotation`	`-a`	Annotation file or directory	(required)
`--format`	`-f`	Annotation format	`auto`
`--domain`	`-d`	Application scenario	`general`
`--sample`	`-s`	Sample size	all
`--output`	`-o`	Output directory	`./qc_report`
`--formats`	-	Report formats	`txt json`
`--no-visual`	-	Disable visualization	false

Scenarios

general: General purpose
road: Autonomous driving (small objects, occlusions)
industrial: Industrial inspection (micro-defects)
security: Security monitoring

Output Example

╔══════════════════════════════════════════╗
║           Image Annotation QC              ║
╠══════════════════════════════════════════╣
║  Images: 1000   Annotations: 4523         ║
║  Errors: 127    Quality Score: 85.5       ║
║  Accuracy: 97.2%                          ║
╚══════════════════════════════════════════╝

Output Files

After QC completes, the following files are generated in the annotation directory:

annotations/
├── qc_report/
│   ├── qc_report.txt      # Text report
│   ├── qc_report.json     # JSON detailed report
│   ├── qc_report.xlsx     # Excel report (optional)
│   └── visual/            # Visualization images
│       ├── frame_001_qc.png
│       └── frame_002_qc.png

Error Types

Error	Description	Weight
Missing	Obvious targets not labeled	1.0
Wrong	Annotation region doesn't match target	0.8
Offset	Box edge deviates >10px	0.5
Too Large	Box significantly larger than target	0.3
Too Small	Box doesn't fully contain target	0.5
Duplicate	Same object annotated multiple times	0.5
Label Error	Wrong class label	0.8

Quality Score

Score = 100 × (annotations - errors×weight) / annotations

Grading:
- 90+: Excellent
- 80-89: Good
- 70-79: Pass
- \x3C70: Fail

Installation

pip install Pillow openpyxl

As Module

from qc_tool import AnnotationQC

qc = AnnotationQC('images/', 'annotations/', format='labelme')
data = qc.load_annotations()
result = qc.check_annotations(data)
qc.generate_report(result, formats=['txt', 'json'])
qc.visualize_errors(result)
print(f"Quality Score: {result.quality_score}")

安全使用建议

This package appears to be what it says: a local image/annotation QC tool. Before running: (1) review and, if needed, run the script on a small sample dataset to confirm behavior; (2) note the script will create a qc_report/ folder under the annotation path and write text/json/xlsx and visualization PNGs; (3) install Pillow and openpyxl from PyPI if you trust that source; (4) if your images or annotations are sensitive, run the tool in an isolated environment since it will read those files locally; and (5) if you want extra assurance, skim the remainder of qc_tool.py (not fully shown in the truncated listing) to confirm there are no unexpected network operations or calls to external services.

功能分析

Type: OpenClaw Skill Name: image-annotation-qc Version: 1.0.7 The skill bundle is a legitimate image annotation quality control tool. The primary script, scripts/qc_tool.py, implements standard logic for validating COCO, YOLO, VOC, and LabelMe formats, including bounding box and polygon checks, duplicate detection, and report generation using Pillow and openpyxl. No evidence of data exfiltration, malicious execution, or prompt injection was found.

能力评估

✓ Purpose & Capability

Name/description match the delivered artifacts: a Python script (qc_tool.py), sample annotation and report files, and SKILL.md describing how to run the QC. Required runtime is only python3 and the declared Python libraries (Pillow, openpyxl) which are reasonable for visualization and Excel output.

✓ Instruction Scope

SKILL.md instructs the agent/user to run the included script against local image and annotation directories and to install Pillow/openpyxl. The instructions and script operate on local files, write reports to a qc_report/ folder, and do not direct data to external endpoints or attempt to read unrelated system paths.

✓ Install Mechanism

There is no automated install spec; SKILL.md suggests pip installing Pillow and openpyxl. That is a standard, proportional dependency for image drawing and Excel output and does not involve arbitrary remote code downloads beyond PyPI.

✓ Credentials

The skill declares no required environment variables, credentials, or config paths. The script only accesses provided image/annotation directories and writes output under qc_report/ — the access requested is proportional to the stated purpose.

✓ Persistence & Privilege

The skill is not marked always:true and does not request persistent platform-level privileges. It creates output in the annotation directory (normal behavior) and does not attempt to modify other skills or global agent configuration.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install image-annotation-qc
安装完成后，直接呼叫该 Skill 的名称或使用 /image-annotation-qc 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v1.0.7

Remove billing, now completely free

v1.0.6

Fix security: remove embedded credentials

v1.0.5

Add SkillPay billing support

v1.0.4

Rebuild

v1.0.3

Revert to English description

v1.0.2

更口语化的描述

v1.0.1

Remove billing module, fix security warning

v1.0.0

Initial release

元数据

Slug image-annotation-qc

版本 1.0.7

许可证 —

累计安装 1

当前安装数 1

历史版本数 8

常见问题