Camelot is a Python library that can help you extract tables from PDFs.
- 📊 Five parsers —
lattice(ruled tables),stream(whitespace), the text-alignmentnetwork/hybrid, and the optional neuralml(Table Transformer) for hard borderless tables — plusflavor="auto"to pick one for you. - 🤖 Borderless & scanned — the optional
mlbackend (pip install "camelot-py[ml]") recovers structure that the heuristic parsers can't on borderless tables; add[ocr]to read scanned / image-only PDFs with no text layer. - 🧠 Vector + raster line detection —
engine="combined"unions the PDF's native vector ruled lines with OpenCV detection, so faintly-ruled tables are still found. - 🐼 pandas output — every table is a
DataFrame, ready for analysis. - 📤 Many export formats — CSV, JSON, Excel, HTML, Markdown, and SQLite.
- 📐 Quality metrics — accuracy, whitespace, and a confidence score per table; drop noise with
TableList.filter(...). - 🧩 Multi-page tables — stitch continuations across pages with
stack_contiguous(). - 🎛️ Highly configurable — table areas/regions, column separators, text processing, and more.
- 🔌 Flexible input — a file path, URL, raw
bytes, or any binary file-like object. - 🖥️ CLI included —
camelot lattice file.pdf, etc. - 📦 Light install — the default pdfium backend is bundled, with no system dependencies.
Extract tables from PDFs in just a few lines of code:
Try it yourself in our interactive quickstart notebook.
Or check out a simple example using this pdf.
>>> import camelot
>>> tables = camelot.read_pdf('foo.pdf')
>>> tables
<TableList n=1>
>>> tables.export('foo.csv', f='csv', compress=True) # json, excel, html, markdown, sqlite
>>> tables[0]
<Table shape=(7, 7)>
>>> tables[0].parsing_report
{
'accuracy': 99.02,
'whitespace': 12.24,
'order': 1,
'page': 1
}
>>> tables[0].to_csv('foo.csv') # to_json, to_excel, to_html, to_markdown, to_sqlite
>>> tables[0].df # get a pandas DataFrame!
| Cycle Name | KI (1/km) | Distance (mi) | Percent Fuel Savings | |||
|---|---|---|---|---|---|---|
| Improved Speed | Decreased Accel | Eliminate Stops | Decreased Idle | |||
| 2012_2 | 3.30 | 1.3 | 5.9% | 9.5% | 29.2% | 17.4% |
| 2145_1 | 0.68 | 11.2 | 2.4% | 0.1% | 9.5% | 2.7% |
| 4234_1 | 0.59 | 58.7 | 8.5% | 1.3% | 8.5% | 3.3% |
| 2032_2 | 0.17 | 57.8 | 21.7% | 0.3% | 2.7% | 1.2% |
| 4171_1 | 0.07 | 173.9 | 58.1% | 1.6% | 2.1% | 0.5% |
Camelot also comes packaged with a command-line interface!
Refer to the QuickStart Guide to quickly get started with Camelot, extract tables from PDFs and explore some basic options.
Tip: Visit the parser-comparison-notebook to get an overview of all the packed parsers and their features.
Note: The built-in parsers need a text-based PDF (as Tabula explains, "If you can click and drag to select text in your table in a PDF viewer, then your PDF is text-based"). For scanned / image-only PDFs, install the neural backend with OCR — pip install "camelot-py[ml,ocr]" — and use camelot.read_pdf(..., flavor="ml"): the model reads the structure from the page image and OCR supplies the text.
You can check out some frequently asked questions here.
| Your PDF | Use | Why |
|---|---|---|
| Ruled tables (lines between cells) | flavor="lattice" (default) |
Deterministic; detects the grid from the ruled lines. engine="combined" also catches faint vector rules. |
| Borderless tables (whitespace-separated) | flavor="network" or "stream" |
Text-alignment / whitespace heuristics — fast, no extra dependencies. |
| Borderless tables, best quality | flavor="ml" (pip install "camelot-py[ml]") |
A Table Transformer model recovers structure heuristics can't — on FinTabNet it roughly doubles borderless TEDS vs network/hybrid. Heavier (PyTorch); opt-in. |
| Scanned / image-only PDFs (no text layer) | flavor="ml" + pip install "camelot-py[ml,ocr]" |
Structure from the model, text from OCR. |
| Mixed / not sure | flavor="auto" |
Picks lattice or network per page. |
The ml backend keeps Camelot honest: the model only supplies the table
structure, while cell text comes from the PDF's own text layer (or OCR
for scans) — so it never invents or alters a value.
- Configurability: Camelot gives you control over the table extraction process with tweakable settings.
- Metrics: You can discard bad tables based on metrics like accuracy and whitespace, without having to manually look at each table.
- Output: Each table is extracted into a pandas DataFrame, which seamlessly integrates into ETL and data analysis workflows. You can also export tables to multiple formats, which include CSV, JSON, Excel, HTML, Markdown, and Sqlite.
See comparison with similar libraries and tools.
Camelot's default image-conversion backend is pdfium, which ships as a wheel — so a plain install needs no system dependencies. The optional ghostscript and poppler backends require additional dependencies.
uv is a fast Python package and project manager. To add Camelot to a project:
uv add camelot-pyOr to install it into the current environment:
uv pip install camelot-pypip install "camelot-py"conda is the package manager for the Anaconda distribution:
conda install -c conda-forge camelot-pygit clone https://github.com/camelot-dev/camelot.git
cd camelot
uv pip install "." # or: pip install "."pip install "camelot-py[ml]" # neural flavor='ml' (Table Transformer; pulls PyTorch)
pip install "camelot-py[ocr]" # OCR text source for scanned PDFs (use with [ml])
pip install "camelot-py[ml,ocr]" # both — borderless + scanned
pip install "camelot-py[plot]" # matplotlib debug plotsThe core install stays light: [ml]/[ocr] are imported lazily, so a plain
import camelot never loads PyTorch or OCR.
The documentation is available at http://camelot-py.readthedocs.io/.
The Contributor's Guide has detailed information about contributing issues, documentation, code, and tests.
Camelot uses Semantic Versioning. For the available versions, see the tags on this repository. For the changelog, you can check out the releases page.
This project is licensed under the MIT License, see the LICENSE file for details.
The documentation theme is licensed under a seperate BSD-like License, see the LICENSE file for details.