drjones/astraea
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Finish: checkbox detection (U+F0A8) + X filling, comprehensive intake (19 Qs), scan all pages

Browse Source
This commit is contained in:
drjones
2026-09-05 15:28:11 -07:00
parent eadce36742
commit 82d0e5b5cb
3 changed files with 59 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
detect_fields.py
View File

@@ -85,6 +85,30 @@ def detect_underscores(pno, page):
return fields
def detect_checkboxes(pno, page):
"""Find checkbox glyphs (U+F0A8) and their labels. Type 'checkbox'."""
fields = []
words = page.get_text("words")
groups = {}
for w in words:
x0, y0, x1, y1, word = w[0], w[1], w[2], w[3], w[4]
key = int(y0 / 5) * 5
groups.setdefault(key, []).append((x0, x1, y0, word))
for y in sorted(groups):
ws = sorted(groups[y], key=lambda t: t[0])
for i, (x0, x1, y0, word) in enumerate(ws):
if "\uf0a8" in word:
label_words = []
for wx0, wx1, _, w in ws[i + 1:]:
if "\uf0a8" in w:
break
label_words.append(w)
label = " ".join(label_words).strip()
fields.append({"page": pno, "x": x0, "y": y0, "type": "checkbox",
"label": label})
return fields
def main():
all_fields = {}
for fn in sorted(os.listdir(FORMS_DIR)):
@@ -99,9 +123,9 @@ def main():
continue
fields = []
seen_keys = set()
for pno in range(min(len(d), 3)): # caption + first 2 pages
lines = lines_with_pos(d[pno])
for pno in range(min(len(d), 12)): # caption + all pages
if pno == 0:
lines = lines_with_pos(d[pno])
for f in detect_caption(lines):
f["page"] = 0
if f["key"] not in seen_keys:
@@ -110,6 +134,9 @@ def main():
for f in detect_underscores(pno, d[pno]):
f["key"] = "f_" + str(len(fields))
fields.append(f)
for f in detect_checkboxes(pno, d[pno]):
f["key"] = "cb_" + str(len(fields))
fields.append(f)
d.close()
all_fields[fid] = fields
out = "/opt/astraea/forms/fieldmap.json"