hocr-spec-python

Validation of hOCR close to the specs

• Rationale
• Installation
• Command line interface
• API example

Rationale

hOCR is a flavor of HTML for encoding the results of Optical Character Recognition (OCR) engines. It is supported by most OCR engines, such as tesseract, ocropus/ocropy and kraken.

The hOCR specifications is at the same time very simple (hOCR is just HTML) and hard to implement, due to its terseness and lack of up-to-date code samples. This project aims to implement the rules defined by the specs from the ground up to serve as a validation tool and reference implementation. It is meant to help hOCR implementers and support tools like hocr-tools.

Installation

Use pip:

# System-wide:
sudo pip install [--user] hocr-spec
# For current user:
pip install --user hocr-spec

From source:

git clone https://github.com/kba/hocr-spec-python
cd hocr-spec-python
# System-wide:
sudo python setup.py install
# For current user:
python setup.py install --user

Command line interface

usage: hocr-spec [-h] [--format {text,bool,ansi,xml}]
                 [--profile {relaxed,standard}]
                 [--implicit_capabilities CAPABILITY]
                 [--skip-check {attributes,classes,metadata,properties}]
                 [--parse-strict] [--silent]
                 sources [sources ...]

positional arguments:
  sources               hOCR file to check or '-' to read from STDIN

optional arguments:
  -h, --help            show this help message and exit
  --format {text,bool,ansi,xml}, -f {text,bool,ansi,xml}
                        Report format
  --profile {relaxed,standard}, -p {relaxed,standard}
                        Validation profile
  --implicit_capabilities CAPABILITY, -C CAPABILITY
                        Enable this capability. Use '*' to enable all
                        capabilities. In addition to the 'ocr*' classes, you
                        can use ['ocrp_dir', 'ocrp_font', 'ocrp_lang',
                        'ocrp_nlp', 'ocrp_poly']
  --skip-check {attributes,classes,metadata,properties}, -X {attributes,classes,metadata,properties}
                        Skip one check
  --parse-strict        Parse HTML with less tolerance for errors
  --silent, -s          Don't produce any output but signal success with exit
                        code.

API example

from hocr_spec import HocrValidator

validator = HocrValidator()
report = validator.validate('/path/to/sample.hocr')
print(report.format('xml'))
# <report valid='false'>...</report>