mirror of
https://github.com/docling-project/docling-eval.git
synced 2026-05-17 13:10:47 +00:00
Christoph Auer
693c22445f
* feat: preserve GraphCellLabel and handle merges in to_value GraphCells Signed-off-by: Christoph Auer <cau@zurich.ibm.com> * fix: GraphCell deduplication with merge groups Signed-off-by: Christoph Auer <cau@zurich.ibm.com> * feat: Review-bundle app builder Signed-off-by: Christoph Auer <cau@zurich.ibm.com> * Update lock Signed-off-by: Christoph Auer <cau@zurich.ibm.com> * Docs and updates for review-bundle-builder Signed-off-by: Christoph Auer <cau@zurich.ibm.com> * fixes for review-bundle-builder and graph-cell merge mapping Signed-off-by: Christoph Auer <cau@zurich.ibm.com> * Eliminate need for page_imgs in cvat_dataset_preannotated, remove json_groundtruth Signed-off-by: Christoph Auer <cau@zurich.ibm.com> * refactor: modernize CVAT dataset builder and improve path handling Replace legacy CVAT annotation processing code (~1400 lines) with modern convert_cvat_folder_to_docling() converter in CvatDatasetBuilder. This significantly simplifies the codebase and aligns with the modern CVAT converter architecture. - Remove legacy annotation parsing and document creation methods - Use convert_cvat_folder_to_docling() for CVAT-to-Docling conversion - Improve path handling in CvatPreannotationBuilder for moved datasets - Remove unused original and original_prediction storage in json_dataset_joiner - Update test data annotations to match new converter output BREAKING CHANGE: CvatDatasetBuilder now requires modern CVAT folder structure and uses convert_cvat_folder_to_docling() internally. Signed-off-by: Christoph Auer <cau@zurich.ibm.com> * feat: combine subset datasets with tags and improve image format support - Refactor cvat_deliveries_to_hf to build single combined dataset with subset tags - Add tags field to DatasetRecord for subset tracking - Add page counting to annotation task creation - Support multipage TIFF and additional image formats (BMP, GIF) - Add configurable JSON directory names to pipeline - Fix caption/footnote target detection logic in CVAT converter Signed-off-by: Christoph Auer <cau@zurich.ibm.com> * Update review bundle README Signed-off-by: Christoph Auer <cau@zurich.ibm.com> * feat: improve label mapping robustness and CVAT delivery flexibility - Use safe label mapping access (.get()) to prevent KeyError exceptions - Filter unsupported labels using DEFAULT_EXPORT_LABELS as safeguard - Add custom directory name support for CVAT deliveries via CLI flags - Preserve original filenames in staging directory - Refactor DeliveryExportKind enum to cleaner value names Signed-off-by: Christoph Auer <cau@zurich.ibm.com> --------- Signed-off-by: Christoph Auer <cau@zurich.ibm.com>