TypeError: The values of index must be integers #2

hopegithub · 2023-12-04T04:54:48Z

hopegithub
Dec 4, 2023

我运行了作者提供的examle中无表格线示例，但报错，我查看我的输入没有问题，是一张表格图片
代码：

# -*- encoding: utf-8 -*-
# @Author: SWHL
# @Contact: [email protected]
from pathlib import Path

from lineless_table_rec import LinelessTableRecognition

engine = LinelessTableRecognition()

img_path = "../test11.jpg"
table_str, elapse = engine(img_path)

print(table_str)
print(elapse)

with open(f"{Path(img_path).stem}.html", "w", encoding="utf-8") as f:
    f.write(table_str)

print("ok")

报错：

WARNING:root:Traceback (most recent call last):
  File "D:\anaconda\envs\paddleocr\lib\site-packages\lineless_table_rec\main.py", line 54, in __call__
    polygons, slct_logi = self.infer(input_info)
  File "D:\anaconda\envs\paddleocr\lib\site-packages\lineless_table_rec\main.py", line 105, in infer
    slct_logi_feat, slct_dets_feat, slct_output_dets = self.det_process(
  File "D:\anaconda\envs\paddleocr\lib\site-packages\lineless_table_rec\lineless_table_process.py", line 26, in __call__
    dets, keep, logi, cr = ctdet_4ps_decode(
  File "D:\anaconda\envs\paddleocr\lib\site-packages\lineless_table_rec\lineless_table_process.py", line 107, in ctdet_4ps_decode
    scores, inds, clses, ys, xs = _topk(heat, K=K)
  File "D:\anaconda\envs\paddleocr\lib\site-packages\lineless_table_rec\lineless_table_process.py", line 206, in _topk
    topk_inds = _gather_feat(topk_inds.reshape(batch, -1, 1), topk_ind).reshape(
  File "D:\anaconda\envs\paddleocr\lib\site-packages\lineless_table_rec\lineless_table_process.py", line 251, in _gather_feat
    feat = _gather(feat, 1, ind)
  File "D:\anaconda\envs\paddleocr\lib\site-packages\lineless_table_rec\lineless_table_process.py", line 277, in _gather
    raise TypeError("The values of index must be integers")
TypeError: The values of index must be integers

Traceback (most recent call last):
  File "D:\工作\checkreport_table\testcode\lineless.py", line 13, in <module>
    print(table_str)
KeyboardInterrupt

Process finished with exit code -1073741510 (0xC000013A: interrupted by Ctrl+C)

请问是什么原因导致的呢

Answered by SWHL

Dec 12, 2023

如果可以翻墙的话，建议去Google Colab上试试，上边有小伙伴已经反映Colab上没有问题了

如果不能翻墙，建议去百度的AI Studio上试试，我在AI Studio上建了一个测试项目表格识别测试

有些怀疑是Windows平台的原因。

如果你那里在linux上测试没有问题的话，方便给我提供一下，你出错的运行环境文件requirements.txt，我这里排查一下原因。

View full answer

nissansz · 2023-12-10T01:34:12Z

nissansz
Dec 10, 2023

解决了吗？也遇到了这个错误

\wired_table_rec>python test.py
Traceback (most recent call last):
File "test.py", line 8, in
table_str, elapse = table_rec(img_path)
File "D:\Python3.8.8\lib\site-packages\wired_table_rec\main.py", line 45, in call
polygons = self.table_line_rec(img)
File "D:\Python3.8.8\lib\site-packages\wired_table_rec\table_line_rec.py", line 37, in call
polygons = self.postprocess(pred)
File "D:\Python3.8.8\lib\site-packages\wired_table_rec\table_line_rec.py", line 97, in postprocess
bbox, _ = bbox_decode(hm[:, 0:1, :, :], c2v, reg=reg, K=self.K)
File "D:\Python3.8.8\lib\site-packages\wired_table_rec\utils_table_line_rec.py", line 17, in bbox_decode
scores, inds, clses, ys, xs = _topk(heat, K=K)
File "D:\Python3.8.8\lib\site-packages\wired_table_rec\utils_table_line_rec.py", line 91, in _topk
topk_inds = _gather_feat(topk_inds.reshape(batch, -1, 1), topk_ind).reshape(
File "D:\Python3.8.8\lib\site-packages\wired_table_rec\utils_table_line_rec.py", line 134, in _gather_feat
feat = _gather_np(feat, 1, ind)
File "D:\Python3.8.8\lib\site-packages\wired_table_rec\utils_table_line_rec.py", line 160, in _gather_np
raise TypeError("The values of index must be integers")
TypeError: The values of index must be integers

0 replies

SWHL · 2023-12-11T00:21:33Z

SWHL
Dec 11, 2023
Maintainer

请尝试用tests目录下的表格图像试试，查看是否报同样的错误？

如果不报错误，说明是提供图像的问题，请方便的话，提供测试的图，便于我这里复现。

0 replies

nissansz · 2023-12-11T00:28:37Z

nissansz
Dec 11, 2023

所有图片都是这个报错

0 replies

nissansz · 2023-12-11T00:35:42Z

nissansz
Dec 11, 2023

系统是win10

0 replies

SWHL · 2023-12-11T01:12:47Z

SWHL
Dec 11, 2023
Maintainer

我这里测试的这张图，并没有报这个错误。

测试代码：

from pathlib import Path
from wired_table_rec import WiredTableRecognition


table_rec = WiredTableRecognition()

img_path = "tests/test_files/wired/row_span.png"
table_str, elapse = table_rec(img_path)
print(table_str)
print(elapse)

with open(f"{Path(img_path).stem}.html", "w", encoding="utf-8") as f:
    f.write(table_str)

输出结果：

<html><body><table><tr><td rowspan="2">名称</td><td rowspan="2">产量（吨）</td><td colspan="2">环比</td></tr><tr><td>增长量（吨）</td><td>增长率（%)</td></tr><tr><td>荔枝</td><td>11</td><td></td><td>10</td></tr><tr><td>芒果</td><td></td><td></td><td>-10</td></tr><tr><td>香蕉</td><td></td><td></td><td>20</td></tr></table></body></html>
1.4739068750059232

如果想要我这里帮助排查问题，请给出详细运行代码、运行环境、python版本、所用图像。

0 replies

nissansz · 2023-12-11T01:17:10Z

nissansz
Dec 11, 2023

代码就是你上面的，图片也是你代码里的，就报错，python3.8.8

!!!
Traceback (most recent call last):
File "F:/pycharm2020.2/RapidStructure-0.0.0/TableStructureRec-main/test.py", line 8, in
table_str, elapse = table_rec(img_path)
File "F:\pycharm2020.2\RapidStructure-0.0.0\TableStructureRec-main\wired_table_rec\main.py", line 45, in call
polygons = self.table_line_rec(img)
File "F:\pycharm2020.2\RapidStructure-0.0.0\TableStructureRec-main\wired_table_rec\table_line_rec.py", line 37, in call
polygons = self.postprocess(pred)
File "F:\pycharm2020.2\RapidStructure-0.0.0\TableStructureRec-main\wired_table_rec\table_line_rec.py", line 97, in postprocess
bbox, _ = bbox_decode(hm[:, 0:1, :, :], c2v, reg=reg, K=self.K)
File "F:\pycharm2020.2\RapidStructure-0.0.0\TableStructureRec-main\wired_table_rec\utils_table_line_rec.py", line 17, in bbox_decode
scores, inds, clses, ys, xs = _topk(heat, K=K)
File "F:\pycharm2020.2\RapidStructure-0.0.0\TableStructureRec-main\wired_table_rec\utils_table_line_rec.py", line 91, in _topk
topk_inds = _gather_feat(topk_inds.reshape(batch, -1, 1), topk_ind).reshape(
File "F:\pycharm2020.2\RapidStructure-0.0.0\TableStructureRec-main\wired_table_rec\utils_table_line_rec.py", line 134, in _gather_feat
feat = _gather_np(feat, 1, ind)
File "F:\pycharm2020.2\RapidStructure-0.0.0\TableStructureRec-main\wired_table_rec\utils_table_line_rec.py", line 160, in _gather_np
raise TypeError("The values of index must be integers")
TypeError: The values of index must be integers
!!!

0 replies

SWHL · 2023-12-11T01:20:22Z

SWHL
Dec 11, 2023
Maintainer

起码要让我这里能复现报错，才能知道是什么问题。

你这总是说报错报错，我也表示无能为力。

0 replies

nissansz · 2023-12-11T01:22:58Z

nissansz
Dec 11, 2023

代码就是你上面的，图片也是你代码里的，所有图片都是同样的报错，python3.8.8

TypeError: The values of index must be integers

0 replies

SWHL · 2023-12-11T01:26:15Z

SWHL
Dec 11, 2023
Maintainer

尝试python3.10 试试，如果也不行的话，那就不是我这里问题了。

0 replies

nissansz · 2023-12-11T01:46:17Z

nissansz
Dec 11, 2023

用colab运行没有报错

0 replies

nissansz · 2023-12-11T01:48:37Z

nissansz
Dec 11, 2023

用倾斜表格测试，识别结果就不对了

0 replies

qqeip · 2023-12-12T09:18:05Z

qqeip
Dec 12, 2023

都报错用样例的图片 3.7.6 3.10.10都一样
if index.dtype != np.dtype("int_"):

      raise TypeError("The values of index must be integers")

E TypeError: The values of index must be integers

E:\xx\site-packages\wired_table_rec-0.0.2-py3.10.egg\wired_table_rec\utils_table_line_rec.py:160: TypeError

2 replies

SWHL Dec 12, 2023
Maintainer

如果可以翻墙的话，建议去Google Colab上试试，上边有小伙伴已经反映Colab上没有问题了

如果不能翻墙，建议去百度的AI Studio上试试，我在AI Studio上建了一个测试项目表格识别测试

有些怀疑是Windows平台的原因。

如果你那里在linux上测试没有问题的话，方便给我提供一下，你出错的运行环境文件requirements.txt，我这里排查一下原因。

Answer selected by SWHL

nissansz Dec 12, 2023

有没有表格区域检测onnx模型？怎么使用？

qqeip · 2023-12-12T15:16:18Z

qqeip
Dec 12, 2023

 [RapidAI/RapidStructure] RapidLayout+RapidTable整合demo (PR #11)

…

---Original--- From: ***@***.***> Date: Tue, Dec 12, 2023 19:02 PM To: ***@***.***>; Cc: ***@***.******@***.***>; Subject: Re: [RapidAI/TableStructureRec] TypeError: The values of index mustbe integers (Discussion #2) 有没有表格区域检测onnx模型？怎么使用？ — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: ***@***.***>

1 reply

nissansz Dec 12, 2023

准确率不太够

qqeip · 2023-12-13T00:05:43Z

qqeip
Dec 13, 2023

Python 3.10.10 WIN10
requirements.txt

1 reply

SWHL Dec 13, 2023
Maintainer

建议把numpy升级到最新试试
我手头没有win10的设备，暂时不能尝试。。。。

qqeip · 2023-12-13T00:21:24Z

qqeip
Dec 13, 2023

aistudio上可以运行但速度有点慢要32秒
可能WIN10上不支持我用pip install wired_table_rec也不行

名称	产量（吨）	环比
名称	产量（吨）	增长量（吨）	增长率（%)
荔枝	11		10
芒果			-10
香蕉			20

32.92959946603514

1 reply

SWHL Dec 13, 2023
Maintainer

速度慢比较正常，AI Studio CPU差一些，后续有很多提速空间。比如用OpenVINO, onnxruntime-gpu版等等

nissansz · 2023-12-13T00:43:14Z

nissansz
Dec 13, 2023

有没有完整前处理python代码？类似这些处理

https://medium.com/@Gekko_lab/make-your-ocr-results-more-accurate-part-ii-preprocessing-3d212ae16191

2 replies

qqeip Dec 13, 2023

cls det模型就是做这事的详见paddleocr 或rapidocr

nissansz Dec 13, 2023

好像不同。我想先前处理整页文档，不是检测文本行

qqeip · 2023-12-13T01:00:09Z

qqeip
Dec 13, 2023

E:\OCR\TableStructureRec\wired_table_rec\utils_table_line_rec.py
def _gather_np(data, dim, index):里把判断异常的两行注释掉就可以运行了

if index.dtype != np.dtype("int_"):

#     raise TypeError("The values of index must be integers")

python demo_wired.py

名称	产量（吨）	环比
名称	产量（吨）	增长量（吨）	增长率（%)
荔枝	11		10
芒果			-10
香蕉			20

2.3827803004533052

16 replies

SWHL Dec 13, 2023
Maintainer

@nissansz
这个建议重新开一个Discussion，已经超出该Discussion的范围

nissansz Dec 13, 2023

如何想可视化画出polygons，要怎么加代码？

main.py

    polygons = self.table_line_rec(img)
    if polygons is None:
        logging.warning("polygons is None.")
        return "", 0.0

    print(polygons)
    import cv2
    # 绘制四边形框
    cv2.polylines(img, polygons, isClosed=True, color=(0, 255, 0), thickness=2)

nissansz Dec 13, 2023

写成这样，可以看到单元格了。但是有的png图片显示成黑底了
print(polygons)
import cv2
import numpy as np
# 重新整理polygons的形状
polygons_int32 = polygons.astype(np.int32)
# 绘制四边形框
cv2.polylines(img, polygons_int32 , isClosed=True, color=(0, 255, 0), thickness=2)
# 显示绘制结果
cv2.imshow('Polygon', img)

明显的边框，有时识别不出来，导致识别结果错误

nissansz Dec 13, 2023

多出很多空单元格

nissansz Dec 13, 2023

倾斜表格的框是对的，但是整合成表格时出错了，不知道要怎么判断，更新，

TypeError: The values of index must be integers #2

Replies: 17 comments · 23 replies

SWHL Dec 11, 2023 Maintainer

SWHL Dec 11, 2023 Maintainer

SWHL Dec 11, 2023 Maintainer

SWHL Dec 11, 2023 Maintainer

SWHL Dec 12, 2023 Maintainer

SWHL Dec 13, 2023 Maintainer

SWHL Dec 13, 2023 Maintainer

if index.dtype != np.dtype("int_"):

SWHL Dec 13, 2023 Maintainer

Replies: 17 comments 23 replies

SWHL
Dec 11, 2023
Maintainer

SWHL
Dec 11, 2023
Maintainer

SWHL
Dec 11, 2023
Maintainer

SWHL
Dec 11, 2023
Maintainer

SWHL Dec 12, 2023
Maintainer

SWHL Dec 13, 2023
Maintainer

SWHL Dec 13, 2023
Maintainer

SWHL Dec 13, 2023
Maintainer