output_files.md 10.8 KB
Newer Older
1
# MinerU 输出文件说明
2

3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
## 概览

`mineru` 命令执行后,除了输出主要的 markdown 文件外,还会生成多个辅助文件用于调试、质检和进一步处理。这些文件包括:

- **可视化调试文件**:帮助用户直观了解文档解析过程和结果
- **结构化数据文件**:包含详细的解析数据,可用于二次开发

下面将详细介绍每个文件的作用和格式。

## 可视化调试文件

### 布局分析文件 (layout.pdf)

**文件命名格式**`{原文件名}_layout.pdf`

**功能说明**
19

20
21
22
23
24
- 可视化展示每一页的布局分析结果
- 每个检测框右上角的数字表示阅读顺序
- 使用不同背景色块区分不同类型的内容块

**使用场景**
25

26
27
28
- 检查布局分析是否正确
- 确认阅读顺序是否合理
- 调试布局相关问题
icecraft's avatar
icecraft committed
29

Sidney233's avatar
Sidney233 committed
30
![layout 页面示例](../images/layout_example.png)
icecraft's avatar
icecraft committed
31

32
33
### 文本片段文件 (spans.pdf)

34
35
> [!NOTE]
> 仅适用于 pipeline 后端
36
37

**文件命名格式**`{原文件名}_spans.pdf`
icecraft's avatar
icecraft committed
38

39
**功能说明**
40

41
42
43
44
- 根据 span 类型使用不同颜色线框标注页面内容
- 用于质量检查和问题排查

**使用场景**
45

46
47
48
- 快速排查文本丢失问题
- 检查行内公式识别情况
- 验证文本分割准确性
icecraft's avatar
icecraft committed
49

Sidney233's avatar
Sidney233 committed
50
![span 页面示例](../images/spans_example.png)
icecraft's avatar
icecraft committed
51

52
53
54
55
## 结构化数据文件

### 模型推理结果 (model.json)

56
57
> [!NOTE]
> 仅适用于 pipeline 后端
icecraft's avatar
icecraft committed
58

59
60
61
**文件命名格式**`{原文件名}_model.json`

#### 数据结构定义
62

icecraft's avatar
icecraft committed
63
64
65
66
67
```python
from pydantic import BaseModel, Field
from enum import IntEnum

class CategoryType(IntEnum):
68
69
70
71
72
73
74
75
76
77
78
79
80
81
    """内容类别枚举"""
    title = 0               # 标题
    plain_text = 1          # 文本
    abandon = 2             # 包括页眉页脚页码和页面注释
    figure = 3              # 图片
    figure_caption = 4      # 图片描述
    table = 5               # 表格
    table_caption = 6       # 表格描述
    table_footnote = 7      # 表格注释
    isolate_formula = 8     # 行间公式
    formula_caption = 9     # 行间公式的标号
    embedding = 13          # 行内公式
    isolated = 14           # 行间公式
    text = 15               # OCR 识别结果
82

icecraft's avatar
icecraft committed
83
class PageInfo(BaseModel):
84
    """页面信息"""
icecraft's avatar
icecraft committed
85
86
87
88
89
    page_no: int = Field(description="页码序号,第一页的序号是 0", ge=0)
    height: int = Field(description="页面高度", gt=0)
    width: int = Field(description="页面宽度", ge=0)

class ObjectInferenceResult(BaseModel):
90
    """对象识别结果"""
icecraft's avatar
icecraft committed
91
    category_id: CategoryType = Field(description="类别", ge=0)
92
    poly: list[float] = Field(description="四边形坐标,格式为 [x0,y0,x1,y1,x2,y2,x3,y3]")
icecraft's avatar
icecraft committed
93
    score: float = Field(description="推理结果的置信度")
94
95
    latex: str | None = Field(description="LaTeX 解析结果", default=None)
    html: str | None = Field(description="HTML 解析结果", default=None)
96

icecraft's avatar
icecraft committed
97
class PageInferenceResults(BaseModel):
98
99
100
    """页面推理结果"""
    layout_dets: list[ObjectInferenceResult] = Field(description="页面识别结果")
    page_info: PageInfo = Field(description="页面元信息")
101

102
# 完整的推理结果
icecraft's avatar
icecraft committed
103
104
105
inference_result: list[PageInferenceResults] = []
```

106
107
108
#### 坐标系统说明

`poly` 坐标格式:`[x0, y0, x1, y1, x2, y2, x3, y3]`
109

110
111
112
- 分别表示左上、右上、右下、左下四点的坐标
- 坐标原点在页面左上角

Sidney233's avatar
Sidney233 committed
113
![poly 坐标示意图](../images/poly.png)
icecraft's avatar
icecraft committed
114

115
#### 示例数据
icecraft's avatar
icecraft committed
116
117
118
119
120
121
122
123

```json
[
    {
        "layout_dets": [
            {
                "category_id": 2,
                "poly": [
124
125
126
127
                    99.1906967163086, 100.3119125366211,
                    730.3707885742188, 100.3119125366211,
                    730.3707885742188, 245.81326293945312,
                    99.1906967163086, 245.81326293945312
icecraft's avatar
icecraft committed
128
129
130
131
132
133
134
135
136
137
138
139
140
                ],
                "score": 0.9999997615814209
            }
        ],
        "page_info": {
            "page_no": 0,
            "height": 2339,
            "width": 1654
        }
    }
]
```

141
142
### VLM 输出结果 (model_output.txt)

143
144
> [!NOTE]
> 仅适用于 VLM 后端
145
146

**文件命名格式**`{原文件名}_model_output.txt`
icecraft's avatar
icecraft committed
147

148
#### 文件格式说明
icecraft's avatar
icecraft committed
149

150
151
- 使用 `----` 分割每一页的输出结果
- 每页包含多个以 `<|box_start|>` 开头、`<|md_end|>` 结尾的文本块
icecraft's avatar
icecraft committed
152

153
#### 字段含义
icecraft's avatar
icecraft committed
154

155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
| 标记 | 格式 | 说明 |
|------|---|------|
| 边界框 | `<\|box_start\|>x0 y0 x1 y1<\|box_end\|>` | 四边形坐标(左上、右下两点),页面缩放至 1000×1000 后的坐标值 |
| 类型标记 | `<\|ref_start\|>type<\|ref_end\|>` | 内容块类型标识 |
| 内容 | `<\|md_start\|>markdown内容<\|md_end\|>` | 该块的 Markdown 内容 |

#### 支持的内容类型

```json
{
    "text": "文本",
    "title": "标题", 
    "image": "图片",
    "image_caption": "图片描述",
    "image_footnote": "图片脚注",
    "table": "表格",
    "table_caption": "表格描述", 
    "table_footnote": "表格脚注",
    "equation": "行间公式"
}
```

#### 特殊标记

- `<|txt_contd|>`:出现在文本末尾,表示该文本块可与后续文本块连接
- 表格内容采用 `otsl` 格式,需转换为 HTML 才能在 Markdown 中渲染

### 中间处理结果 (middle.json)

**文件命名格式**`{原文件名}_middle.json`

#### 顶层结构

| 字段名 | 类型 | 说明 |
|--------|------|------|
| `pdf_info` | `list[dict]` | 每一页的解析结果数组 |
| `_backend` | `string` | 解析模式:`pipeline``vlm` |
| `_version_name` | `string` | MinerU 版本号 |

#### 页面信息结构 (pdf_info)

| 字段名 | 说明 |
|--------|------|
| `preproc_blocks` | PDF 预处理后的未分段中间结果 |
| `layout_bboxes` | 布局分割结果,包含布局方向和边界框,按阅读顺序排序 |
| `page_idx` | 页码,从 0 开始 |
| `page_size` | 页面的宽度和高度 `[width, height]` |
| `_layout_tree` | 布局树状结构 |
| `images` | 图片块信息列表 |
| `tables` | 表格块信息列表 |
| `interline_equations` | 行间公式块信息列表 |
| `discarded_blocks` | 需要丢弃的块信息 |
| `para_blocks` | 分段后的内容块结果 |

#### 块结构层次

```
一级块 (table | image)
└── 二级块
    └── 行 (line)
        └── 片段 (span)
```
icecraft's avatar
icecraft committed
217

218
#### 一级块字段
icecraft's avatar
icecraft committed
219

220
221
222
223
224
| 字段名 | 说明 |
|--------|------|
| `type` | 块类型:`table``image` |
| `bbox` | 块的矩形框坐标 `[x0, y0, x1, y1]` |
| `blocks` | 包含的二级块列表 |
icecraft's avatar
icecraft committed
225

226
#### 二级块字段
icecraft's avatar
icecraft committed
227

228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
| 字段名 | 说明 |
|--------|------|
| `type` | 块类型(详见下表) |
| `bbox` | 块的矩形框坐标 |
| `lines` | 包含的行信息列表 |

#### 二级块类型

| 类型 | 说明 |
|------|------|
| `image_body` | 图像本体 |
| `image_caption` | 图像描述文本 |
| `image_footnote` | 图像脚注 |
| `table_body` | 表格本体 |
| `table_caption` | 表格描述文本 |
| `table_footnote` | 表格脚注 |
| `text` | 文本块 |
| `title` | 标题块 |
| `index` | 目录块 |
| `list` | 列表块 |
| `interline_equation` | 行间公式块 |

#### 行和片段结构

**行 (line) 字段**
- `bbox`:行的矩形框坐标
- `spans`:包含的片段列表

**片段 (span) 字段**
- `bbox`:片段的矩形框坐标
- `type`:片段类型(`image``table``text``inline_equation``interline_equation`
- `content` | `img_path`:文本内容或图片路径

#### 示例数据
icecraft's avatar
icecraft committed
262
263
264
265
266

```json
{
    "pdf_info": [
        {
267
            "preproc_blocks": [...],
icecraft's avatar
icecraft committed
268
269
            "layout_bboxes": [
                {
270
                    "layout_bbox": [52, 61, 294, 731],
icecraft's avatar
icecraft committed
271
272
273
274
275
                    "layout_label": "V",
                    "sub_layout": []
                }
            ],
            "page_idx": 0,
276
            "page_size": [612.0, 792.0],
icecraft's avatar
icecraft committed
277
278
279
            "para_blocks": [
                {
                    "type": "text",
280
                    "bbox": [52, 61.956024169921875, 294, 82.99800872802734],
icecraft's avatar
icecraft committed
281
282
                    "lines": [
                        {
283
                            "bbox": [52, 61.956024169921875, 294, 72.0000228881836],
icecraft's avatar
icecraft committed
284
285
                            "spans": [
                                {
286
287
                                    "bbox": [54.0, 61.956024169921875, 296.2261657714844, 72.0000228881836],
                                    "content": "dependent on the service headway...",
icecraft's avatar
icecraft committed
288
289
290
291
292
293
294
295
296
297
                                    "type": "text",
                                    "score": 1.0
                                }
                            ]
                        }
                    ]
                }
            ]
        }
    ],
298
    "_backend": "pipeline",
icecraft's avatar
icecraft committed
299
300
    "_version_name": "0.6.1"
}
301
```
302

303
304
305
306
307
308
309
310
311
### 内容列表 (content_list.json)

**文件命名格式**`{原文件名}_content_list.json`

#### 功能说明

这是一个简化版的 `middle.json`,按阅读顺序平铺存储所有可读内容块,去除了复杂的布局信息,便于后续处理。

#### 内容类型
312

313
314
315
316
317
318
| 类型 | 说明 |
|------|------|
| `image` | 图片 |
| `table` | 表格 |
| `text` | 文本/标题 |
| `equation` | 行间公式 |
319

320
#### 文本层级标识
321

322
通过 `text_level` 字段区分文本层级:
323

324
325
326
327
-`text_level``text_level: 0`:正文文本
- `text_level: 1`:一级标题
- `text_level: 2`:二级标题
- 以此类推...
328

329
330
331
332
333
#### 通用字段

所有内容块都包含 `page_idx` 字段,表示所在页码(从 0 开始)。

#### 示例数据
334
335
336
337
338

```json
[
    {
        "type": "text",
339
340
        "text": "The response of flow duration curves to afforestation",
        "text_level": 1,
341
342
343
        "page_idx": 0
    },
    {
344
345
        "type": "text", 
        "text": "Abstract",
346
347
348
349
350
351
        "text_level": 2,
        "page_idx": 0
    },
    {
        "type": "image",
        "img_path": "images/a8ecda1c69b27e4f79fce1589175a9d721cbdc1cf78b4cc06a015f3746f6b9d8.jpg",
352
        "img_caption": ["Fig. 1. Annual flow duration curves..."],
353
354
355
356
357
        "img_footnote": [],
        "page_idx": 1
    },
    {
        "type": "equation",
358
359
        "img_path": "images/181ea56ef185060d04bf4e274685f3e072e922e7b839f093d482c29bf89b71e8.jpg", 
        "text": "$$Q_{\\%} = f(P) + g(T)$$",
360
361
362
363
364
365
        "text_format": "latex",
        "page_idx": 2
    },
    {
        "type": "table",
        "img_path": "images/e3cb413394a475e555807ffdad913435940ec637873d673ee1b039e3bc3496d0.jpg",
366
367
368
        "table_caption": ["Table 2 Significance of the rainfall and time terms"],
        "table_footnote": ["indicates that the rainfall term was significant..."],
        "table_body": "<html><body><table>...</table></body></html>",
369
370
371
        "page_idx": 5
    }
]
372
373
374
375
376
377
378
379
380
381
```

## 总结

以上文件为 MinerU 的完整输出结果,用户可根据需要选择合适的文件进行后续处理:

- **模型输出**:使用原始输出(model.json、model_output.txt)
- **调试和验证**:使用可视化文件(layout.pdf、spans.pdf) 
- **内容提取**:使用简化文件(*.md、content_list.json)
- **二次开发**:使用结构化文件(middle.json)