Summary
- 图片OCR识别工具,提取图片中的文字内容,支持中英文混合识别、代码截图优化、表格数据提取,让纯文本模型也能理解图片。
- Use when:
- - 提取图片中的文字 extract text from images
- - 代码截图识别 recognize code screenshots
- - 表格图片转结构化数据…
wulaosiji/skills
图片OCR识别工? Use when: - 提取图片中的文字 extract text from images - 代码截图识别 recognize code screenshots - 表格图片转结构化数据 convert table images to structured data - 让文本模型理解图片? - 批量图片文字提取 batch image text extraction - 文档扫描件OCR OCR scanned documents Part of UniqueClub toolkit. Learn more: https://uniqueclub.ai
npx skills add wulaosiji/skills --skill image-ocr
Related neighbors and high-traction skills in the same topics — useful to compare before installing.
Process and manipulate images using ImageMagick. Supports resizing, format conversion, batch pr…
9.6K installsGraalVM Native Image expert that adds native image support to Java applications, builds the pro…
8.7K installsGenerate or edit raster images when the task benefits from AI-created bitmap visuals such as ph…
4.3K installsCreates a complete EC2 Image Builder pipeline that builds a custom AMI with pre-installed softw…
4K installsBuild Azure managed images and Azure Compute Gallery images with Packer. Use when creating cust…
2.3K installs>- Use when the user wants to orchestrate defect image generation with NVIDIA Cosmos AnomalyGen…
1.9K installsOther skills from wulaosiji/skills · top by installs.
npx skills add wulaosiji/skills
Declared targets from SKILL.md / docs. Unmarked agents are not listed — the skill may still install via the CLI.
main
Files included with this skill beyond the listing page.
SKILL.md
6,945 B
SUMMARY.md
686 B
将图片中的文字提取出来,让不支持视觉的模型(如 Qwen-Coder)也能"看懂"图片内容。
Chinese:
English:
| 引擎 | 特点 | 适用场景 |
|---|---|---|
| PaddleOCR | 离线、免费、中英支持好 | 日常使用、代码识别 |
| Baidu OCR | API、更准确、支持手写 | 手写体、高要求场景 |
| Tencent OCR | API、多语言 | 多语言需求 |
# config.json
{
"ocr_engine": "paddle",
"language": "ch_sim",
"save_temp": false
}
from skills.image_ocr.scripts.ocr import recognize
result = recognize("/path/to/image.png")
print(result.text)
# PaddleOCR(推荐)
pip3 install paddleocr -i https://pypi.tuna.tsinghua.edu.cn/simple
# 或完整版
pip3 install paddlepaddle paddleocr
# 基本用法
python3 skills/image-ocr/scripts/ocr.py /path/to/image.png
# 带选项
python3 skills/image-ocr/scripts/ocr.py /path/to/image.png --format markdown --save
from skills.image_ocr.ocr_engine import OCREngine
ocr = OCREngine(engine="paddle")
result = ocr.recognize("image.png")
print(result.text) # 完整文字
print(result.blocks) # 文字块详情
print(result.confidence) # 置信度
这是图片中的文字内容:
第一行文字
第二行文字
## 图片内容识别结果
def hello(): print("Hello World")
表格数据:
| 列1 | 列2 |
|-----|-----|
| A | B |
{
"text": "完整文字",
"blocks": [
{"text": "第一块", "confidence": 0.98, "position": [x1,y1,x2,y2]},
{"text": "第二块", "confidence": 0.95, "position": [x1,y1,x2,y2]}
],
"language": "ch_sim"
}
针对代码截图的特殊处理:
# 自动检测代码并格式化
ocr_result = ocr.recognize(image_path)
if is_code_screenshot(ocr_result):
formatted = format_as_code(ocr_result)
Features:
0→O,1→l){
"ocr_engine": "paddle",
"baidu_api_key": "your_key",
"baidu_secret_key": "your_secret",
"language": "ch_sim",
"save_temp": false
}
| 参数 | 选项 | 说明 |
|---|---|---|
ocr_engine |
paddle/baidu/tencent | OCR引擎选择 |
language |
chsim/en/chtra | 语言:简体中文/英文/繁体中文 |
save_temp |
true/false | 是否保存临时文件 |
if message.type == "image":
image_path = download_image(message.file_key)
ocr_result = run_skill("image-ocr", image_path)
text_prompt = f"用户发了一张图片,内容是:\n{ocr_result}\n请根据以上内容回答..."
response = query_model("qwen-coder", text_prompt)
用户:帮我识别这张图里的代码
[图片]
Agent:我来识别一下...
[调用 ocr skill]
Agent:图片里的代码是:...
Q: PaddleOCR安装失败?
# 使用 Conda 安装
conda install paddlepaddle -c conda-forge
pip install paddleocr
Q: 识别准确率不高?
language 配置Q: 支持多语言吗?
| Skill | Relationship | Use Case |
|---|---|---|
| content-extractor | 内容提取 | 提取图片内容到文本 |
| document-hub | 文档处理 | 将OCR结果生成文档 |
| 格式处理 | PDF扫描件OCR | |
| long-form-writer | 内容加工 | 基于OCR内容写作 |
| md-to-wechat | 输出转换 | 转换为公众号格式 |
skills/image-ocr/
├── SKILL.md
├── config.json # 配置文件
├── ocr_engine.py # OCR引擎封装
├── post_processor.py # 后处理(代码格式化等)
└── scripts/
└── ocr.py # CLI入口
Part of the UniqueClub toolkit - a collection of skills for AI-powered content creation and automation.