How to Extract Tables From PDF With OCR (Free Methods)

Extracting table data from PDFs is one of the most frustrating tasks in data work. Whether you’re a researcher pulling numbers from published reports, an analyst importing financial data, or an accountant reconciling invoices, getting structured data out of PDFs without manual retyping is a huge time saver.

---

Types of PDF Tables

Before choosing a tool, identify what type of PDF you’re dealing with:

1. Native digital PDF — the PDF was created from software. Text is real characters. Table lines may or may not be present as vector graphics.

2. Scanned/Image PDF — the page is an image. The “table” is just pixels. Requires OCR before data can be extracted.

---

Method 1: Camelot (Python — Best for Native PDFs)

Camelot is purpose-built for table extraction from native PDFs. It’s consistently the most accurate free tool.

import camelot

# Lattice mode — for tables with visible borders/lines
tables = camelot.read_pdf('annual_report.pdf', flavor='lattice', pages='1-3')

# Stream mode — for tables without borders (space-separated columns)
tables = camelot.read_pdf('financial_data.pdf', flavor='stream', pages='all')

# Check accuracy score
print(tables[0].parsing_report)

# Export to various formats
tables.export('output.csv', f='csv')
tables.export('output.xlsx', f='excel')
tables.export('output.json', f='json')

# Access as pandas DataFrame
df = tables[0].df
print(df.head())

---

Method 2: Tabula (GUI + Python)

Tabula is a desktop app (free, open-source) that lets you draw boxes around tables visually and extract them.

1. Download and open Tabula
2. Upload your PDF
3. Draw a selection box around the table
4. Click Preview & Export → CSV or Excel

Also available as Python library:

import tabula

# Extract all tables from a page
tables = tabula.read_pdf('report.pdf', pages='2', multiple_tables=True)

# Export to Excel
tabula.convert_into('report.pdf', 'output.xlsx', output_format='xlsx', pages='all')

---

Method 3: pdfplumber (Python — Precise Positioning)

pdfplumber gives you detailed control over PDF parsing and is excellent for complex multi-column layouts.

import pdfplumber

with pdfplumber.open('document.pdf') as pdf:
    for page_num, page in enumerate(pdf.pages, 1):
        tables = page.extract_tables()
        for table_num, table in enumerate(tables, 1):
            print(f"Page {page_num}, Table {table_num}:")
            for row in table:
                print(row)

---

Method 4: Google Sheets Import (No-Code)

For one-off extractions without any coding:

1. Upload the PDF to Google Drive
2. Right-click → Open with Google Docs
3. Google’s OCR extracts the content including table structure
4. Copy the table → paste into Google Sheets
5. Clean up any formatting issues

Alternatively: Google Sheets → File → Import → select PDF (works on some digital PDFs).

---

Method 5: Scanned PDFs — OCR + Table Extraction

For scanned PDFs, you need OCR first:

import pdf2image
import pytesseract
import pandas as pd
from PIL import Image

def extract_table_from_scanned_pdf(pdf_path, page_num=1):
    # Convert PDF page to image
    images = pdf2image.convert_from_path(pdf_path, dpi=300)
    image = images[page_num - 1]
    
    # OCR with table data output format
    data = pytesseract.image_to_data(image, output_type=pytesseract.Output.DATAFRAME)
    
    # Filter confident results
    data = data[data.conf > 60].dropna(subset=['text'])
    return data

# Better: Use pytesseract's HOCR output for table reconstruction
html = pytesseract.image_to_pdf_or_hocr(image, extension='hocr')

For complex scanned tables, ABBYY FineReader (paid) significantly outperforms free tools.

---

Comparison of Table Extraction Tools

Tool	Native PDFs	Scanned PDFs	Borders Required	Output Formats
Camelot	★★★★★	✗	Lattice only	CSV, Excel, JSON
Tabula	★★★★☆	✗	No	CSV, Excel
pdfplumber	★★★★☆	✗	No	Raw Python objects
Google Sheets	★★★☆☆	★★★☆☆	No	Google Sheets
ABBYY FineReader	★★★★★	★★★★★	No	All formats

---

Troubleshooting Common Issues

Tables spanning multiple pages: Use pages='all' in Camelot/Tabula and merge DataFrames with pd.concat().

Merged cells: These are the hardest to handle automatically. Use camelot’s copy_text=['v'] option to fill merged cells.

Mixed text and numbers: Always cast numeric columns after extraction with pd.to_numeric(col, errors='coerce').

For a comprehensive guide to PDF data extraction, the PyPDF community documentation is an excellent reference.