Benjamin Admin be7f5f1872 feat: Sprint 2 — TrOCR ONNX, PP-DocLayout, Model Management ...

D2: TrOCR ONNX export script (printed + handwritten, int8 quantization)
D3: PP-DocLayout ONNX export script (download or Docker-based conversion)
B3: Model Management admin page (PyTorch vs ONNX status, benchmarks, config)
A4: TrOCR ONNX service with runtime routing (auto/pytorch/onnx via TROCR_BACKEND)
A5: PP-DocLayout ONNX detection with OpenCV fallback (via GRAPHIC_DETECT_BACKEND)
B4: Structure Detection UI toggle (OpenCV vs PP-DocLayout) with class color coding
C3: TrOCR-ONNX.md documentation
C4: OCR-Pipeline.md ONNX section added
C5: mkdocs.yml nav updated, optimum added to requirements.txt

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

2026-03-23 09:53:02 +01:00

..

data

…

mail

…

migrations

feat: Sprint 1 — IPA hardening, regression framework, ground-truth review

2026-03-23 09:21:27 +01:00

models

…

policies

…

routes

…

services

feat: Sprint 2 — TrOCR ONNX, PP-DocLayout, Model Management

2026-03-23 09:53:02 +01:00

tests

feat: Sprint 2 — TrOCR ONNX, PP-DocLayout, Model Management

2026-03-23 09:53:02 +01:00

admin_api.py

…

config.py

…

country_metadata.py

…

cv_box_detect.py

feat: run shading-based box detection alongside line detection

2026-03-16 08:12:52 +01:00

cv_cell_grid.py

fix: word_boxes auch fuer breite Spalten (Full-Page OCR) speichern

2026-03-11 20:41:29 +01:00

cv_color_detect.py

Fix bullet overlap disambiguation + raise red threshold to 90

2026-03-20 18:21:00 +01:00

cv_doclayout_detect.py

feat: Sprint 2 — TrOCR ONNX, PP-DocLayout, Model Management

2026-03-23 09:53:02 +01:00

cv_graphic_detect.py

feat: Sprint 2 — TrOCR ONNX, PP-DocLayout, Model Management

2026-03-23 09:53:02 +01:00

cv_layout.py

feat: box-aware column detection — exclude box content from global columns

2026-03-16 18:42:46 +01:00

cv_ocr_engines.py

feat: Sprint 1 — IPA hardening, regression framework, ground-truth review

2026-03-23 09:21:27 +01:00

cv_preprocessing.py

fix: swap 90°/270° rotation direction in orientation detection

2026-03-17 16:39:15 +01:00

cv_review.py

fix: _group_words_into_lines nach cv_ocr_engines.py verschieben

2026-03-09 15:24:56 +01:00

cv_vocab_pipeline.py

feat: Words-First Grid Builder (bottom-up alternative zu cell_grid_v2)

2026-03-12 06:46:05 +01:00

cv_vocab_types.py

Vertical zone split: detect divider lines and create independent sub-zones

2026-03-20 16:38:12 +01:00

cv_words_first.py

fix: sort word_boxes in reading order (Y-grouped, then X-sorted)

2026-03-17 10:41:30 +01:00

dsfa_corpus_ingestion.py

…

dsfa_rag_api.py

…

eh_pipeline.py

…

eh_templates.py

…

embedding_client.py

…

full_compliance_pipeline.py

…

full_reingestion.py

…

github_crawler.py

…

grid_editor_api.py

fix: hard-filter OCR words inside detected graphic regions

2026-03-22 10:18:23 +01:00

handwriting_htr_api.py

…

hybrid_search.py

…

hybrid_vocab_extractor.py

…

hyde.py

…

legal_corpus_api.py

…

legal_corpus_ingestion.py

…

legal_corpus_robust.py

…

legal_templates_ingestion.py

…

main.py

feat: add Excel-like grid editor for OCR overlay (Kombi mode step 6)

2026-03-14 23:41:03 +01:00

metrics_db.py

…

migrate_rag_chunks.py

…

minio_storage.py

…

nibis_ingestion.py

…

nru_worksheet_generator.py

…

ocr_labeling_api.py

…

ocr_pipeline_api.py

Add Ground Truth regression test system for OCR pipeline

2026-03-18 13:46:48 +01:00

ocr_pipeline_auto.py

refactor: split ocr_pipeline_api.py (5426 lines) into 8 modules

2026-03-18 08:42:00 +01:00

ocr_pipeline_common.py

refactor: split ocr_pipeline_api.py (5426 lines) into 8 modules

2026-03-18 08:42:00 +01:00

ocr_pipeline_geometry.py

Invalidate grid_editor_result when exclude regions change

2026-03-19 09:19:09 +01:00

ocr_pipeline_ocr_merge.py

refactor: split ocr_pipeline_api.py (5426 lines) into 8 modules

2026-03-18 08:42:00 +01:00

ocr_pipeline_overlays.py

Preserve alphabetic marker columns, broaden junk filter, enable IPA in grid

2026-03-18 11:08:23 +01:00

ocr_pipeline_postprocess.py

refactor: split ocr_pipeline_api.py (5426 lines) into 8 modules

2026-03-18 08:42:00 +01:00

ocr_pipeline_regression.py

feat: Sprint 1 — IPA hardening, regression framework, ground-truth review

2026-03-23 09:21:27 +01:00

ocr_pipeline_rows.py

refactor: split ocr_pipeline_api.py (5426 lines) into 8 modules

2026-03-18 08:42:00 +01:00

ocr_pipeline_session_store.py

Add Ground Truth regression test system for OCR pipeline

2026-03-18 13:46:48 +01:00

ocr_pipeline_sessions.py

refactor: split ocr_pipeline_api.py (5426 lines) into 8 modules

2026-03-18 08:42:00 +01:00

ocr_pipeline_words.py

refactor: split ocr_pipeline_api.py (5426 lines) into 8 modules

2026-03-18 08:42:00 +01:00

orientation_crop_api.py

feat: auto-detect multi-page spreads and split into sub-sessions

2026-03-17 16:34:06 +01:00

page_crop.py

Fix spine shadow false positives: require dark valley, brightness rise, trim convolution edges

2026-03-19 08:23:50 +01:00

pdf_export.py

…

pdf_extraction.py

…

pipeline_checkpoints.py

…

pyproject.toml

…

qdrant_service.py

…

rag_evaluation.py

…

rbac.py

…

requirements.txt

feat: Sprint 2 — TrOCR ONNX, PP-DocLayout, Model Management

2026-03-23 09:53:02 +01:00

reranker.py

…

self_rag.py

…

storage.py

…

template_sources.py

…

tesseract_vocab_extractor.py

…

training_api.py

…

training_export_service.py

…

trocr_api.py

…

upload_api.py

…

vocab_session_store.py

…

vocab_worksheet_api.py

fix: Edge-Gaps in _split_broad_columns ignorieren + return-Tuple bei leerem Ergebnis

2026-03-07 22:16:29 +01:00

worksheet_cleanup_api.py

…

worksheet_editor_api.py

…

zeugnis_api.py

…

zeugnis_crawler.py

…

zeugnis_models.py

…

zeugnis_seed_data.py

…