Files
Christoph Auer 8fb3a169f6 fix: consistenty and perf improvements (#171)
* Add validation-only mode to cvat_evaluation_pipeline and make output consistent

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* Revise CVAT XML parsing for efficiency

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* Revise CVAT evaluation pipeline for efficiency

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* fix: use LiveText OCR and refine picture validation

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* fix: keep table containers in level‑1 order after conflict resolution

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* fix: respect inline reading order for embedded headers

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* fix: skip cvat samples with oversized ocr pages

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* fix: tighten reading-order validation resilience

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* fix: convert grouped elements skipped by reading order

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* fix: keep conflicted reading-order containers independent

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* fix: skip container merges during cvat conversion

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* fix: guard merge paths that bridge separate containers

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* fix: reconcile doc tree after CVAT rich cell rewiring

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* feat: track OCR usage during CVAT conversion

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

* feat: Add tool to upload CVAT-generated datasets to HF

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>

---------

Signed-off-by: Christoph Auer <cau@zurich.ibm.com>
2025-11-10 11:48:32 +01:00
..
2025-10-08 15:52:56 +02:00