signverod / README.md
ondrs's picture
Update README.md
951d659 verified
metadata
dataset_info:
  features:
    - name: image
      dtype: image
    - name: width
      dtype: float64
    - name: height
      dtype: float64
    - name: objects
      struct:
        - name: id
          sequence: int64
        - name: area
          sequence: float64
        - name: bbox
          sequence:
            sequence: float64
        - name: category
          sequence: int64
    - name: image_id
      dtype: int64
  splits:
    - name: train
      num_bytes: 788454917.485
      num_examples: 2299
  download_size: 512699837
  dataset_size: 788454917.485
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: apache-2.0
tags:
  - signature
  - initials
  - handwritten
  - redaction
  - date
pretty_name: 'SignverOD: A Dataset Signature Object Detection'

Dataset Card for SignverOD

Detecting the presence and location of hand written artifacts such as signatures, dates, initials can be critical for scanned (offline)document processing systems. This capability can support multiple downstream tasks such as signature verification, document tagging and categorization.

SignverOD is a curated dataset of 2576 scanned document images with 7103 bounding box annotations, across 4 categories (signature, initials, redaction, date). SignverOD cover a diverse set of document types including memos, emails, bank cheques, lease agreements and letters, memos, invoices.

Dataset Sources

Original source: https://www.kaggle.com/datasets/victordibia/signverod

Images of documents in this dataset are sourced from 4 main locations and then annotated.

Tobacco800

Tobacco800 is a publicly accessible document image collection with realistic scope and complexity is important to the document image analysis and search community. https://www.kaggle.com/sprytte/tobacco-800-dataset.

NIST.gov Special Database

The NIST.gov structured Forms Database consists of 5,590 pages of binary, black-and-white images of synthesized documents. The documents in this database are 12 different tax forms from the IRS 1040 Package X for the year 1988. https://www.nist.gov/srd/nist-special-database-2

Bank Cheques

The bank cheques dataset is a collection of xx colored images of bank checks. They consist of scanned realistic checks as well as examplar signatures with signatures. https://www.kaggle.com/saifkhichi96/bank-checks-signatures-segmentation-dataset

GSA.gov Lease Documents

GSA provides electronic copies of GSA lease documents for general public viewing. The lease documents are sorted by region and contain, for the most part, GSA Lease Forms and Lease Amendments (LA) from selected GSA leases across the nation. https://www.gsa.gov/real-estate/real-estate-services/leasing/executed-lease-documents

CiItation

If you use this dataset as part of your work or experiments, please consider citing:

@article{DibiaReed2022signverod, 
  author = {Victor, Dibia},
  title = {A Dataset for Handwritten Signature Object Detection in Scanned Documents.},
  year = {2022},
  publisher = {victordibia.com},
  journal = {victordibia.com},
  url = {https://victordibia.com/signverod.pdf}
}

@article{Dibia2022signver,
  author = {Victor, Dibia and Andrew Reed},
  title = {SignVer: A Deep Learning Library for Automatic Offline Signature Verification Tasks},
  year = {2022},
  publisher = {victordibia.com},
  journal = {victordibia.com},
  url = {https://victordibia.com/signver.pdf}
}