RapidAI · SWHL · Mar 31, 2025 · Mar 7, 2025 · Mar 8, 2025 · Mar 8, 2025
diff --git a/.github/workflows/lineless_table_rec.yml b/.github/workflows/lineless_table_rec.yml
@@ -29,11 +29,7 @@ jobs:
         run: |
           pip install -r requirements.txt
           pip install pytest
-
-          wget https://github.com/RapidAI/TableStructureRec/releases/download/v0.0.0/lineless_table_rec_models.zip
-          unzip lineless_table_rec_models.zip
-          mv lineless_table_rec_models/*.onnx lineless_table_rec/models/
-
+          pip install rapidocr
           pytest tests/test_lineless_table_rec.py
 
   GenerateWHL_PushPyPi:
@@ -54,11 +50,7 @@ jobs:
           pip install -r requirements.txt
           python -m pip install --upgrade pip
           pip install wheel get_pypi_latest_version
-
-          wget https://github.com/RapidAI/TableStructureRec/releases/download/v0.0.0/lineless_table_rec_models.zip
-          unzip lineless_table_rec_models.zip
-          mv lineless_table_rec_models/*.onnx lineless_table_rec/models/
-
+          pip install rapidocr
           python setup_lineless.py bdist_wheel "${{ github.ref_name }}"
 
       # - name: Publish distribution 📦 to Test PyPI

diff --git a/.github/workflows/table_cls.yml b/.github/workflows/table_cls.yml
@@ -29,10 +29,6 @@ jobs:
           pip install -r requirements.txt
           pip install pytest beautifulsoup4
 
-          wget https://github.com/RapidAI/TableStructureRec/releases/download/v0.0.0/table_cls_models.zip
-          unzip table_cls_models.zip
-          mv table_cls_models/*.onnx table_cls/models/
-
           pytest tests/test_table_cls.py
 
   GenerateWHL_PushPyPi:
@@ -54,10 +50,6 @@ jobs:
           python -m pip install --upgrade pip
           pip install wheel get_pypi_latest_version
 
-          wget https://github.com/RapidAI/TableStructureRec/releases/download/v0.0.0/table_cls_models.zip
-          unzip table_cls_models.zip
-          mv table_cls_models/*.onnx table_cls/models/
-
           python setup_table_cls.py bdist_wheel "${{ github.ref_name }}"
 
       - name: Publish distribution 📦 to PyPI

diff --git a/.github/workflows/wired_table_rec.yml b/.github/workflows/wired_table_rec.yml
@@ -28,11 +28,7 @@ jobs:
         run: |
           pip install -r requirements.txt
           pip install pytest beautifulsoup4
-
-          wget https://github.com/RapidAI/TableStructureRec/releases/download/v0.0.0/wired_table_rec_models.zip
-          unzip wired_table_rec_models.zip
-          mv wired_table_rec_models/*.onnx wired_table_rec/models/
-
+          pip install rapidocr
           pytest tests/test_wired_table_rec.py
 
   GenerateWHL_PushPyPi:
@@ -53,11 +49,7 @@ jobs:
           pip install -r requirements.txt
           python -m pip install --upgrade pip
           pip install wheel get_pypi_latest_version
-
-          wget https://github.com/RapidAI/TableStructureRec/releases/download/v0.0.0/wired_table_rec_models.zip
-          unzip wired_table_rec_models.zip
-          mv wired_table_rec_models/*.onnx wired_table_rec/models/
-
+          pip install rapidocr
           python setup_wired.py bdist_wheel "${{ github.ref_name }}"
 
       - name: Publish distribution 📦 to PyPI

diff --git a/README.md b/README.md
@@ -15,12 +15,16 @@
 </div>
 
 ### 最近更新
-- **2024.11.22**
-  - 支持单字符匹配方案，需要RapidOCR>=1.4.0
 - **2024.12.25**
     - 补充文档扭曲矫正/去模糊/去阴影/二值化方案，可作为前置处理 [RapidUnDistort](https://github.com/Joker1212/RapidUnWrap)
 - **2025.1.9**
-  - RapidTable支持了 unitable 模型，精度更高支持torch推理，补充测评数据   
+  - RapidTable支持了 unitable 模型，精度更高支持torch推理，补充测评数据
+- **2025.3.30**
+    - 输入输出格式对齐RapidTable
+    - 支持模型自动下载
+    - 增加来自paddle的新表格分类模型
+    - 增加最新PaddleX表格识别模型测评值
+    - 支持 rapidocr 2.0 取消重复ocr检测
 
 ### 简介
 💖该仓库是用来对文档中表格做结构化识别的推理库，包括来自阿里读光有线和无线表格识别模型，llaipython(微信)贡献的有线表格模型，网易Qanything内置表格分类模型等。\
@@ -54,18 +58,19 @@
    Surya-Tabled 使用内置ocr模块，表格模型为行列识别模型，无法识别单元格合并，导致分数较低
 
 | 方法                                                                                                       |    TEDS     | TEDS-only-structure |
-|:---------------------------------------------------------------------------------------------------------|:-----------:|:-------------------:|
-| [surya-tabled(--skip-detect)](https://github.com/VikParuchuri/tabled)                                    |   0.33437   |       0.65865       |
-| [surya-tabled](https://github.com/VikParuchuri/tabled)                                                   |   0.33940   |       0.67103       |
-| [deepdoctection(table-transformer)](https://github.com/deepdoctection/deepdoctection?tab=readme-ov-file) |   0.59975   |       0.69918       |
-| [ppstructure_table_master](https://github.com/PaddlePaddle/PaddleOCR/tree/main/ppstructure)              |   0.61606   |       0.73892       |
-| [ppsturcture_table_engine](https://github.com/PaddlePaddle/PaddleOCR/tree/main/ppstructure)              |   0.67924   |       0.78653       |
-| [StructEqTable](https://github.com/UniModal4Reasoning/StructEqTable-Deploy)                              |   0.67310   |       0.81210       |
-| [RapidTable(SLANet)](https://github.com/RapidAI/RapidTable)                                              |   0.71654   |       0.81067       |
-| table_cls + wired_table_rec v1 + lineless_table_rec                                                      |   0.75288   |       0.82574       |
-| table_cls + wired_table_rec v2 + lineless_table_rec                                                      |   0.77676   |       0.84580       |
-| [RapidTable(SLANet-plus)](https://github.com/RapidAI/RapidTable)                                         | 0.84481 |     0.91369     |
-| [RapidTable(unitable)](https://github.com/RapidAI/RapidTable)                                         | **0.86200** |     **0.91813**     |
+|:---------------------------------------------------------------------------------------------------------|:-----------:|:-----------------:|
+| [surya-tabled(--skip-detect)](https://github.com/VikParuchuri/tabled)                                    |   0.33437   |       0.65865     |
+| [surya-tabled](https://github.com/VikParuchuri/tabled)                                                   |   0.33940   |       0.67103     |
+| [deepdoctection(table-transformer)](https://github.com/deepdoctection/deepdoctection?tab=readme-ov-file) |   0.59975   |       0.69918     |
+| [ppstructure_table_master](https://github.com/PaddlePaddle/PaddleOCR/tree/main/ppstructure)              |   0.61606   |       0.73892     |
+| [ppsturcture_table_engine](https://github.com/PaddlePaddle/PaddleOCR/tree/main/ppstructure)              |   0.67924   |       0.78653     |
+| [StructEqTable](https://github.com/UniModal4Reasoning/StructEqTable-Deploy)                              |   0.67310   |       0.81210     |
+| [RapidTable(SLANet)](https://github.com/RapidAI/RapidTable)                                              |   0.71654   |       0.81067     |
+| table_cls + wired_table_rec v1 + lineless_table_rec                                                      |   0.75288   |       0.82574     |
+| table_cls + wired_table_rec v2 + lineless_table_rec                                                      |   0.77676   |       0.84580     |
+| [PaddleX(SLANetXt+RT-DERT)](https://github.com/PaddlePaddle/PaddleX)                                                                            |   0.79900   |       **0.92222**     |
+| [RapidTable(SLANet-plus)](https://github.com/RapidAI/RapidTable)                                         |   0.84481   |       0.91369     |
+| [RapidTable(unitable)](https://github.com/RapidAI/RapidTable)                                            | **0.86200** |     0.91813     |
 
 ### 使用建议
 wired_table_rec_v2(有线表格精度最高): 通用场景有线表格(论文，杂志，期刊, 收据，单据，账单)
@@ -75,63 +80,93 @@ wired_table_rec_v2 对1500px内大小的图片效果最好，所以分辨率超
 SLANet-plus/unitable (综合精度最高): 文档场景表格(论文，杂志，期刊中的表格)
 
 ### 安装
-
+rapidocr2.0以上版本支持torch,onnx,paddle,openvino等多引擎切换，详情参考[rapidocr文档](https://rapidai.github.io/RapidOCRDocs/main/install_usage/rapidocr/usage/)
 ``` python {linenos=table}
 pip install wired_table_rec lineless_table_rec table_cls
+pip install rapidocr 
 ```
 
 ### 快速使用
-
+> ⚠️注意：在`wired_table_rec/table_cls`>=1.2.0` `lineless_table_rec` > 0.1.0 后，采用同RapidTable完全一致格式的输入输出
 ``` python {linenos=table}
-import os
+from pathlib import Path
 
-from lineless_table_rec import LinelessTableRecognition
-from lineless_table_rec.utils_table_recover import format_html, plot_rec_box_with_logic_info, plot_rec_box
+from wired_table_rec.utils.utils import VisTable
 from table_cls import TableCls
-from wired_table_rec import WiredTableRecognition
-from rapidocr_onnxruntime import RapidOCR 
-
-lineless_engine = LinelessTableRecognition()
-wired_engine = WiredTableRecognition()
-# 默认小yolo模型(0.1s)，可切换为精度更高yolox(0.25s),更快的qanything(0.07s)模型
-table_cls = TableCls() # TableCls(model_type="yolox"),TableCls(model_type="q")
-img_path = f'images/img14.jpg'
-
-cls,elasp = table_cls(img_path)
-if cls == 'wired':
-    table_engine = wired_engine
-else:
-    table_engine = lineless_engine
-
-html, elasp, polygons, logic_points, ocr_res = table_engine(img_path)
-print(f"elasp: {elasp}")
-
-# 使用其他ocr模型
-#ocr_engine =RapidOCR(det_model_path="xxx/det_server_infer.onnx",rec_model_path="xxx/rec_server_infer.onnx")
-#ocr_res, _ = ocr_engine(img_path)
-#html, elasp, polygons, logic_points, ocr_res = table_engine(img_path, ocr_result=ocr_res)
-# output_dir = f'outputs'
-# complete_html = format_html(html)
-# os.makedirs(os.path.dirname(f"{output_dir}/table.html"), exist_ok=True)
-# with open(f"{output_dir}/table.html", "w", encoding="utf-8") as file:
-#     file.write(complete_html)
-# # 可视化表格识别框 + 逻辑行列信息
-# plot_rec_box_with_logic_info(
-#     img_path, f"{output_dir}/table_rec_box.jpg", logic_points, polygons
-# )
-# # 可视化 ocr 识别框
-# plot_rec_box(img_path, f"{output_dir}/ocr_box.jpg", ocr_res)
+from wired_table_rec.main import WiredTableInput, WiredTableRecognition
+from lineless_table_rec.main import LinelessTableInput, LinelessTableRecognition
+from rapidocr import RapidOCR
+
+
+if __name__ == "__main__":
+    # Init
+    wired_input = WiredTableInput()
+    lineless_input = LinelessTableInput()
+    wired_engine = WiredTableRecognition(wired_input)
+    lineless_engine = LinelessTableRecognition(lineless_input)
+    viser = VisTable()
+    # 默认小yolo模型(0.1s)，可切换为精度更高yolox(0.25s),更快的qanything(0.07s)模型或paddle模型(0.03s)
+    table_cls = TableCls()
+    img_path = f"tests/test_files/table.jpg"
+
+    cls, elasp = table_cls(img_path)
+    if cls == "wired":
+        table_engine = wired_engine
+    else:
+        table_engine = lineless_engine
+
+    # 使用RapidOCR输入
+    ocr_engine = RapidOCR()
+    rapid_ocr_output = ocr_engine(img_path, return_word_box=True)
+    ocr_result = list(
+        zip(rapid_ocr_output.boxes, rapid_ocr_output.txts, rapid_ocr_output.scores)
+    )
+    table_results = table_engine(
+        img_path, ocr_result=ocr_result
+    )
+
+    # 使用单字识别
+    # word_results = rapid_ocr_output.word_results
+    # ocr_result = [
+    #     [word_result[2], word_result[0], word_result[1]] for word_result in word_results
+    # ]
+    # table_results = table_engine(
+    #     img_path, ocr_result=ocr_result, enhance_box_line=False
+    # )
+
+    # Save
+    # save_dir = Path("outputs")
+    # save_dir.mkdir(parents=True, exist_ok=True)
+    # 
+    # save_html_path = f"outputs/{Path(img_path).stem}.html"
+    # save_drawed_path = f"outputs/{Path(img_path).stem}_table_vis{Path(img_path).suffix}"
+    # save_logic_path = (
+    #     f"outputs/{Path(img_path).stem}_table_vis_logic{Path(img_path).suffix}"
+    # )
+
+    # Visualize table rec result
+    # vis_imged = viser(
+    #     img_path, table_results, save_html_path, save_drawed_path, save_logic_path
+    # )
+
+
+
+
+
 ```
 
 #### 单字ocr匹配
+
 ```python
 # 将单字box转换为行识别同样的结构)
-from rapidocr_onnxruntime import RapidOCR
-from wired_table_rec.utils_table_recover import trans_char_ocr_res
+from rapidocr import RapidOCR
 img_path = "tests/test_files/wired/table4.jpg"
-ocr_engine =RapidOCR()
-ocr_res, _ = ocr_engine(img_path, return_word_box=True)
-ocr_res = trans_char_ocr_res(ocr_res)
+ocr_engine = RapidOCR()
+rapid_ocr_output = ocr_engine(img_path, return_word_box=True)
+word_results = rapid_ocr_output.word_results
+ocr_result = [
+    [word_result[2], word_result[0], word_result[1]] for word_result in word_results
+]
 ```
 
 #### 表格旋转及透视修正
@@ -177,24 +212,53 @@ for i, res in enumerate(result):
 
 ### 核心参数
 ```python
-wired_table_rec = WiredTableRecognition()
-html, elasp, polygons, logic_points, ocr_res = wired_table_rec(
+# 输入(WiredTableInput/LinelessTableInput)
+@dataclass
+class WiredTableInput:
+    model_type: Optional[str] = "unet" #unet/cycle_center_net
+    model_path: Union[str, Path, None, Dict[str, str]] = None
+    use_cuda: bool = False
+    device: str = "cpu"
+
+@dataclass
+class LinelessTableInput:
+    model_type: Optional[str] = "lore" #lore
+    model_path: Union[str, Path, None, Dict[str, str]] = None
+    use_cuda: bool = False
+    device: str = "cpu"
+
+# 输出(WiredTableOutput/LinelessTableOutput)
+@dataclass
+class WiredTableOutput:
+    pred_html: Optional[str] = None
+    cell_bboxes: Optional[np.ndarray] = None
+    logic_points: Optional[np.ndarray] = None
+    elapse: Optional[float] = None
+
+@dataclass
+class LinelessTableOutput:
+    pred_html: Optional[str] = None
+    cell_bboxes: Optional[np.ndarray] = None
+    logic_points: Optional[np.ndarray] = None
+    elapse: Optional[float] = None
+```
+
+```python
+wired_table_rec = WiredTableRecognition(WiredTableInput())
+table_results = wired_table_rec(
     img, # 图片 Union[str, np.ndarray, bytes, Path, PIL.Image.Image]
     ocr_result, # 输入rapidOCR识别结果，不传默认使用内部rapidocr模型
-    version="v2", #默认使用v2线框模型，切换阿里读光模型可改为v1
     enhance_box_line=True, # 识别框切割增强(关闭避免多余切割，开启减少漏切割)，默认为True
     col_threshold=15, # 识别框左边界x坐标差值小于col_threshold的默认同列
     row_threshold=10, # 识别框上边界y坐标差值小于row_threshold的默认同行
     rotated_fix=True, # wiredV2支持，轻度旋转(-45°~45°)矫正，默认为True
     need_ocr=True, # 是否进行OCR识别, 默认为True
-    rec_again=True,# 是否针对未识别到文字的表格框,进行单独截取再识别,默认为True
 )
-lineless_table_rec = LinelessTableRecognition()
-html, elasp, polygons, logic_points, ocr_res = lineless_table_rec(
+lineless_table_rec = LinelessTableRecognition(LinelessTableInput())
+table_results = lineless_table_rec(
     img, # 图片 Union[str, np.ndarray, bytes, Path, PIL.Image.Image]
     ocr_result, # 输入rapidOCR识别结果，不传默认使用内部rapidocr模型
     need_ocr=True, # 是否进行OCR识别, 默认为True
-    rec_again=True,# 是否针对未识别到文字的表格框,进行单独截取再识别,默认为True
 )
 ```
 
@@ -225,7 +289,7 @@ html, elasp, polygons, logic_points, ocr_res = lineless_table_rec(
 ```mermaid
 flowchart TD
     A[/表格图片/] --> B([表格分类 table_cls])
-    B --> C([有线表格识别 wired_table_rec]) & D([无线表格识别 lineless_table_rec]) --> E([文字识别 rapidocr_onnxruntime])
+    B --> C([有线表格识别 wired_table_rec]) & D([无线表格识别 lineless_table_rec]) --> E([文字识别 rapidocr])
     E --> F[/html结构化输出/]
 ```