TableStructureRec
整理目前开源的最优表格识别模型，完善前后处理，模型转换为ONNX | Organize the currently open-source optimal table recognition models, improve pre-processing and post-processing, and convert the models to ONNX.
Install / Use
/learn @RapidAI/TableStructureRecREADME
English | 简体中文
</div>最近更新
- 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
- 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
- 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。
快速开始 模型评测 使用建议 单字匹配 文档扭曲修正 表格旋转及透视修正 输入参数 常见问题 更新计划
特点
⚡ 快 采用ONNXRuntime作为推理引擎，cpu下单图推理1-7s
🎯 准: 结合表格类型分类模型，区分有线表格，无线表格，任务更细分，精度更高
🛡️ 稳: 不依赖任何第三方训练框架，只依赖必要基础库，避免包冲突
在线演示
效果展示
<div align="center"> <img src="https://github.com/RapidAI/TableStructureRec/releases/download/v0.0.0/demo_img_output.gif" alt="Demo" width="100%" height="100%"> </div>指标结果
TableRecognitionMetric 评测工具 huggingface数据集 modelscope 数据集 Rapid OCR
测试环境: ubuntu 20.04 python 3.10.10 opencv-python 4.10.0.84
注: StructEqTable 输出为 latex，只取成功转换为html并去除样式标签后进行测评
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) | 0.33437 | 0.65865 | | surya-tabled | 0.33940 | 0.67103 | | deepdoctection(table-transformer) | 0.59975 | 0.69918 | | ppstructure_table_master | 0.61606 | 0.73892 | | ppsturcture_table_engine | 0.67924 | 0.78653 | | StructEqTable | 0.67310 | 0.81210 | | RapidTable(SLANet) | 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) | 0.79900 | 0.92222 | | RapidTable(SLANet-plus) | 0.84481 | 0.91369 | | RapidTable(unitable) | 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)
wired_table_rec_v2 对1500px内大小的图片效果最好，所以分辨率超过2000px建议等比缩放一下
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)
版本依赖关系
|库|版本|rapidocr|
|:---|:---|:---|
|wired_table_rec|v1.2.0|rapidocr>1.0.0,<3.0.0|
|lineless_table_rec|v0.1.0|rapidocr>1.0.0,<3.0.0|
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
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>=1.2.0lineless_table_rec` > 0.1.0 后，采用同RapidTable完全一致格式的输入输出
from pathlib import Path
from wired_table_rec.utils.utils import VisTable
from table_cls import TableCls
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
# 将单字box转换为行识别同样的结构)
from rapidocr import RapidOCR
img_path = "tests/test_files/wired/table4.jpg"
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
1.简单背景，小角度场景
最新wiredV2模型自适应小角度旋转
import cv2
img_path = f'tests/test_files/wired/squeeze_error.jpeg'
from wired_table_rec.utils import ImageOrientationCorrector
img_orientation_corrector = ImageOrientationCorrector()
img = cv2.imread(img_path)
img = img_orientation_corrector(img)
cv2.imwrite(f'img_rotated.jpg', img)
2.复杂背景，多表格场景
需要gpu或更高精度场景，请参考项目RapidTableDet
pip install rapid-table-det
import os
import cv2
from rapid_table_det.utils import img_loader, visuallize, extract_table_img
from rapid_table_det.inference import TableDetector
table_det = TableDetector()
img_path = f"tests/test_files/chip.jpg"
result, elapse = table_det(img_path)
img = img_loader(img_path)
extract_img = img.copy()
#可能有多表格
for i, res in enumerate(result):
box = res["box"]
lt, rt, rb, lb = res["lt"], res["rt"], res["rb"], res["lb"]
# 带识别框和左上角方向位置
img = visuallize(img, box, lt, rt, rb, lb)
# 透视变换提取表格图片
wrapped_img = extract_table_img(extract_img.copy(), lt, rt, rb, lb)
# cv2.imwrite(f"{out_dir}/{file_name}-extract-{i}.jpg", wrapped_img)
# cv2.imwrite(f"{out_dir}/{file_name}-visualize.jpg", img)
核心参数
# 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
Related Skills
node-connect
337.7kDiagnose OpenClaw node connection and pairing failures for Android, iOS, and macOS companion apps
frontend-design
83.3kCreate distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, or applications. Generates creative, polished code that avoids generic AI aesthetics.
openai-whisper-api
337.7kTranscribe audio via OpenAI Audio Transcriptions API (Whisper).
commit-push-pr
83.3kCommit, push, and open a PR
