zlei6/ragflow
1
0
Fork 0
mirror of https://github.com/infiniflow/ragflow.git synced 2026-07-28 19:58:11 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix(deepdoc): match table columns within the same page (#17282)

Browse Source
This commit is contained in:
Ehsan
2026-07-24 05:39:57 +03:00
committed by GitHub
parent 1b07bf286d
commit e25b929a38
2 changed files with 176 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
deepdoc/vision/recognizer.py
View File

@@ -238,6 +238,13 @@ class Recognizer:
for i, b in enumerate(boxes):
if box.get("layoutno", "0") != b.get("layoutno", "0"):
continue
# layoutno is f"table-{index}" and resets per page, so the same string
# names a different table on every page. Only accept a candidate that
# shares vertical extent with the box; top/bottom are page-cumulative,
# so this rejects a same-layoutno column from another page whose x range
# happens to be closer.
if min(box["bottom"], b["bottom"]) <= max(box["top"], b["top"]):
continue
dis = min(abs(box["x0"] - b["x0"]), abs(box["x1"] - b["x1"]), abs(box["x0"] + box["x1"] - b["x1"] - b["x0"]) / 2)
if dis < min_dis:
min_i = i