hocr-check

#!/usr/bin/env python

# check the given file for conformance with the hOCR format spec

import argparse
import sys

from lxml import html

################################################################
# misc library code
################################################################

TEST_COUNTER = 0


def test_ok(v, msg):
    global TEST_COUNTER
    TEST_COUNTER += 1
    if not v:
        sys.stderr.write("not ")
    sys.stderr.write("ok " + str(TEST_COUNTER) + " - " + msg + "\n")


# node properties


def get_prop(node, name):
    title = node.get('title')
    if not title:
        return None
    props = title.split(';')
    for prop in props:
        (key, args) = prop.split(None, 1)
        if key == name:
            return args
    return None


def get_bbox(node):
    bbox = get_prop(node, 'bbox')
    if not bbox:
        return None
    return tuple([int(x) for x in bbox.split()])


# rectangle properties


def intersect(u, v):
    # intersection of two rectangles
    r = (max(u[0], v[0]), max(u[1], v[1]), min(u[2], v[2]), min(u[3], v[3]))
    return r


def area(u):
    # area of a rectangle
    return max(0, u[2] - u[0]) * max(0, u[3] - u[1])


def overlaps(u, v):
    # predicate: do the two rectangles overlap?
    return area(intersect(u, v)) > 0


def relative_overlap(u, v):
    m = max(area(u), area(v))
    i = area(intersect(u, v))
    return float(i) / m


def mostly_nonoverlapping(boxes, significant_overlap=0.2):
    for i in range(len(boxes)):
        for j in range(i + 1, len(boxes)):
            if relative_overlap(boxes[i], boxes[j]) > significant_overlap:
                return 0
    return 1


################################################################
# main
################################################################

parser = argparse.ArgumentParser(
    description=("Check the given file for conformance with the hOCR "
                 "format spec")
)
parser.add_argument(
    "file",
    help="hOCR file to check",
    type=argparse.FileType('r'),
    nargs='?',
    default=sys.stdin)
parser.add_argument(
    "-o",
    "--nooverlap",
    help="Disable the overlap checks",
    action="store_true")
args = parser.parse_args()

doc = html.parse(args.file)

################################################################
# XML structure checks
################################################################

# check for presence of meta information
test_ok(
    doc.xpath("//meta[@name='ocr-system']") != [],
    "//meta[@name='ocr-system']")
test_ok(
    doc.xpath("//meta[@name='ocr-capabilities']") != [],
    "//meta[@name='ocr-capabilities']")

# check for presence of page
test_ok(doc.xpath("//*[@class='ocr_page']") != [], "has a page")

# check that lines are inside pages
lines = doc.xpath("//*[@class='ocr_line']")
for line_idx, line in enumerate(lines):
    test_ok(
        line.xpath("./ancestor::*[@class='ocr_page']"),
        "ocr_line %2d in an ocr_page" % (line_idx))

# check that pars are inside pages
pars = doc.xpath("//*[@class='ocr_par']")
for par_idx, par in enumerate(pars):
    test_ok(
        par.xpath("./ancestor::*[@class='ocr_page']"),
        "ocr_par %2d in an ocr_page" % (par_idx))

# check that careas are inside pages
careas = doc.xpath("//*[@class='ocr_carea']")
for carea_idx, carea in enumerate(careas):
    test_ok(
        carea.xpath("./ancestor::*[@class='ocr_page']"),
        "ocr_carea %2d in an ocr_page" % (carea_idx))

################################################################
# geometric checks
################################################################

if not args.nooverlap:
    for page in doc.xpath("//*[@class='ocr_page']"):
        # check lines
        objs = page.xpath("//*[@class='ocr_line']")
        line_bboxes = [get_bbox(obj) for obj in objs if get_prop(obj, 'bbox')]
        test_ok(
            mostly_nonoverlapping(line_bboxes), 'mostly_nonoverlapping/line')
        # check paragraphs
        objs = page.xpath("//*[@class='ocr_par']")
        par_bboxes = [get_bbox(obj) for obj in objs if get_prop(obj, 'bbox')]
        test_ok(mostly_nonoverlapping(par_bboxes), 'mostly_nonoverlapping/par')
        # check careas
        objs = page.xpath("//*[@class='ocr_carea']")
        carea_bboxes = [get_bbox(obj) for obj in objs if get_prop(obj, 'bbox')]
        test_ok(
            mostly_nonoverlapping(carea_bboxes), 'mostly_nonoverlapping/carea')

################################################################
# TODO
################################################################

# FIXME add many other checks:
# - containment of paragraphs, careas, etc.
# - ocr-capabilities vs. actual tags
# - warn about text outside ocr_ elements
# - check title= attribute format
# - check that only the right attributes are present on the right elements
# - check for unrecognized ocr_ elements
# - check for significant overlaps
# - check that image files are not repeated