img2table

1.4.2 verified Mon Apr 27 auth: no python

img2table is a table identification and extraction library for PDFs and images, based on OpenCV image processing. Current version: 1.4.2. Supports Python 3.9-3.13. Released on PyPI with moderate cadence.

pip install img2table

Common errors

error ModuleNotFoundError: No module named 'paddleocr' ↓

cause PaddleOCR is an extra dependency, not installed by default with img2table.

fix

pip install paddleocr

error ImportError: cannot import name 'PaddleOCR' from 'img2table.ocr' ↓

cause Using wrong casing; correct class name is PaddleOCR (capital O, C, R).

fix

Use: from img2table.ocr import PaddleOCR

error AttributeError: 'Image' object has no attribute 'extract_tables' ↓

cause Incorrect import; Image class is not in top-level package.

fix

Use: from img2table.document import Image

Warnings

breaking In v1.4.0, the PDF backend was migrated from PyMuPDF/fitz to pypdfium2 for license compliance. Existing code expecting fitz will break. ↓

fix No action needed if using Document classes; only affects direct use of PDF library internals.

deprecated The old TesseractOCR class used Tesseract 4.x; future versions may remove support. Recommended to migrate to PaddleOCR or SuryaOCR. ↓

fix Switch to PaddleOCR or SuryaOCR via pip install paddleocr or pip install surya-ocr (separate).

gotcha OCR initialization is heavy; avoid recreating OCR instance per image in loops. Reuse the same OCR object for multiple documents. ↓

fix Create one OCR object and pass it to multiple extract_tables calls.

Imports

OCR

wrong

from img2table.ocr import PaddleOcr

correct

from img2table.ocr import PaddleOCR, TesseractOCR

Wrong casing; class is PaddleOCR (capital O, C, R).

Document
wrong
```
from img2table import Image
```
correct
```
from img2table.document import Image, PDF
```
Image and PDF are classes inside img2table.document, not top-level.

Quickstart

Extract tables from an image using PaddleOCR.

import os
from img2table.document import Image
from img2table.ocr import PaddleOCR

# Use environment variable for API key if needed
ocr = PaddleOCR(lang='en', api_key=os.environ.get('PADDLE_OCR_KEY', ''))

img = Image(src='table.png')
tables = img.extract_tables(ocr=ocr)
print(tables)