Benjamin_Boenisch/breakpilot-lehrer
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(ocr-pipeline): cells = result, no post-processing content shuffling

Browse Source 
The cell grid IS the result. Each cell stays at its detected position.

Removed _split_comma_entries and _attach_example_sentences from the
pipeline — they were shuffling content between rows/columns, causing
"Mäuse" to appear in a separate row, "stand..." to move to Example,
and "Ei" to disappear.

Now: cells → _cells_to_vocab_entries (1:1 row mapping) →
_fix_character_confusion → _fix_phonetic_brackets → done.

Also lowered pixel-density threshold from 2% to 0.5% for the cell-OCR
fallback so small text like "Ei" is not filtered out.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Benjamin Admin
2026-03-02 09:41:30 +01:00
parent e3f939a628
commit 72cc77dcf4
2 changed files with 8 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
klausur-service/backend/cv_vocab_pipeline.py
View File

@@ -3186,9 +3186,11 @@ def _ocr_single_cell(
if ocr_img is not None:
crop = ocr_img[cell_y:cell_y + cell_h, cell_x:cell_x + cell_w]
if crop.size > 0:
# Threshold: pixels darker than 180 (on 0-255 grayscale)
# Threshold: pixels darker than 180 (on 0-255 grayscale).
# Use 0.5% to catch even small text like "Ei" (2 chars)
# in an otherwise empty cell.
dark_ratio = float(np.count_nonzero(crop < 180)) / crop.size
_run_fallback = dark_ratio > 0.02
_run_fallback = dark_ratio > 0.005
if _run_fallback:
cell_region = PageRegion(
type=col.type,

13
klausur-service/backend/ocr_pipeline_api.py
View File

@@ -35,8 +35,6 @@ from cv_vocab_pipeline import (
_cells_to_vocab_entries,
_fix_character_confusion,
_fix_phonetic_brackets,
_split_comma_entries,
_attach_example_sentences,
analyze_layout,
analyze_layout_by_words,
build_cell_grid,
@@ -1174,15 +1172,13 @@ async def detect_words(
},
}
# For vocab layout: add post-processed vocab_entries (backwards compat)
# For vocab layout: map cells 1:1 to vocab entries (row→entry).
# No content shuffling — each cell stays at its detected position.
if is_vocab:
entries = _cells_to_vocab_entries(cells, columns_meta)
entries = _fix_character_confusion(entries)
entries = _fix_phonetic_brackets(entries, pronunciation=pronunciation)
entries = _split_comma_entries(entries)
entries = _attach_example_sentences(entries)
word_result["vocab_entries"] = entries
# Also keep "entries" key for backwards compatibility
word_result["entries"] = entries
word_result["entry_count"] = len(entries)
word_result["summary"]["total_entries"] = len(entries)
@@ -1302,14 +1298,13 @@ async def _word_stream_generator(
},
}
# Vocab post-processing
# For vocab layout: map cells 1:1 to vocab entries (row→entry).
# No content shuffling — each cell stays at its detected position.
vocab_entries = None
if is_vocab:
entries = _cells_to_vocab_entries(all_cells, columns_meta)
entries = _fix_character_confusion(entries)
entries = _fix_phonetic_brackets(entries, pronunciation=pronunciation)
entries = _split_comma_entries(entries)
entries = _attach_example_sentences(entries)
word_result["vocab_entries"] = entries
word_result["entries"] = entries
word_result["entry_count"] = len(entries)