---
title: pdfFormDetector
emoji: 📄
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: 6.13.0
app_file: app.py
pinned: false
license: mit
---

# CommonForms — Form Field Detector

Gradio Space que usa o modelo **FFDNet-L** (`jbarrow/FFDNet-L-cpu`) via o
pacote oficial [`commonforms`](<https://pypi.org/project/commonforms/>) para
converter PDFs em formulários preenchíveis, detectando:

- text boxes
- checkboxes (choice buttons)
- signature fields (opcional)

Paper: [CommonForms — arxiv 2509.16506](<https://arxiv.org/abs/2509.16506>)

## Arquivos

| Arquivo           | Propósito                                                      |
|-------------------|----------------------------------------------------------------|
| `app.py`          | UI Gradio + endpoint `/detect`                                 |
| `requirements.txt`| `gradio==6.13.0` + `commonforms`                               |
| `packages.txt`    | `poppler-utils` (necessário pra `pdf2image`)                   |
| `README.md`       | Metadata do Space + docs                                       |

O pacote `commonforms` baixa o ONNX do FFDNet-L do HF Hub no primeiro uso.

## Endpoint

​```python
from gradio_client import Client, handle_file

client = Client("luisabwk/pdfFormDetector")
out_pdf = client.predict(
    pdf_path=handle_file("/caminho/input.pdf"),
    image_size=1600,
    use_signature_fields=False,
    keep_existing_fields=False,
    api_name="/detect",
)
print(out_pdf)  # path para o PDF preenchível
​```

## Parâmetros

- **image_size** (512–2048, default 1600) — maior = mais preciso, mais lento.
- **use_signature_fields** — detecta regiões de assinatura.
- **keep_existing_fields** — preserva widgets AcroForm já presentes.