60 lines
1.6 KiB
Markdown
60 lines
1.6 KiB
Markdown
|
# OCR 重构说明
|
||
|
|
|
||
|
|
## 主要变更
|
||
|
|
|
||
|
|
### 1. 简化的处理器类型
|
||
|
|
- 移除了复杂的本地调用 MinerU 版本
|
||
|
|
- 统一使用 HTTP API 接口
|
||
|
|
- 现在只有 3 种处理器: RapidOCR, MinerU, PaddleX
|
||
|
|
|
||
|
|
### 2. 统一的接口
|
||
|
|
所有处理器都实现 `BaseDocumentProcessor` 接口:
|
||
|
|
- `process_file()` - 处理文件
|
||
|
|
- `check_health()` - 健康检查
|
||
|
|
- `get_service_name()` - 获取服务名
|
||
|
|
|
||
|
|
### 3. 使用方式
|
||
|
|
```python
|
||
|
|
# 新方式 (推荐)
|
||
|
|
from src.plugins.document_processor_factory import DocumentProcessorFactory
|
||
|
|
text = DocumentProcessorFactory.process_file("mineru_ocr", file_path, params)
|
||
|
|
|
||
|
|
# 旧方式 (仍可用)
|
||
|
|
from src.plugins import ocr
|
||
|
|
text = ocr.process_file_mineru(file_path, params)
|
||
|
|
```
|
||
|
|
|
||
|
|
## 迁移说明
|
||
|
|
|
||
|
|
### 代码迁移
|
||
|
|
```python
|
||
|
|
# 旧代码 (仍可用)
|
||
|
|
from src.plugins import ocr
|
||
|
|
text = ocr.process_file_mineru(file_path, params)
|
||
|
|
|
||
|
|
# 新代码 (推荐)
|
||
|
|
from src.plugins.document_processor_factory import DocumentProcessorFactory
|
||
|
|
text = DocumentProcessorFactory.process_file("mineru_ocr", file_path, params)
|
||
|
|
```
|
||
|
|
|
||
|
|
### 处理器类型
|
||
|
|
- `onnx_rapid_ocr`: RapidOCR (CPU)
|
||
|
|
- `mineru_ocr`: MinerU HTTP API (GPU)
|
||
|
|
- `paddlex_ocr`: PaddleX (GPU)
|
||
|
|
|
||
|
|
### 环境变量
|
||
|
|
```bash
|
||
|
|
# MinerU API
|
||
|
|
export MINERU_API_URI="http://localhost:30001"
|
||
|
|
|
||
|
|
# PaddleX
|
||
|
|
export PADDLEX_URI="http://localhost:8080"
|
||
|
|
```
|
||
|
|
|
||
|
|
## 主要改进
|
||
|
|
|
||
|
|
1. **简化架构**: 从4个处理器减少到3个
|
||
|
|
2. **统一接口**: 所有处理器实现相同接口
|
||
|
|
3. **内存优化**: PDF 处理内存占用从 ~800MB 降至 ~150MB
|
||
|
|
4. **错误处理**: 统一异常处理机制
|
||
|
|
5. **图片验证**: 强制要求图片文件必须启用 OCR
|