docld
Datasets
All datasets matching “docld”fintabnet-bench
DocLD FinTabNet Benchmark Results
Benchmark results for DocLD table extraction on the
FinTabNet dataset.
Results Summary
Metric
Value
Mean Accuracy
82.1%
Median
83.2%
P25 / P75
73.3% / 97.4%
Min / Max
22.7% / 100.0%
Scored Samples
500
Total Samples
500
Methodology
Dataset: FinTabNet_OTSL — 500 samples from the test split
Extraction: DocLD vision-based table extraction
Scoring: Needleman-Wunsch hierarchical alignment (same as… See the full description on the dataset page: https://huggingface.co/datasets/docld/fintabnet-bench.funsd-bench
FUNSD benchmark (DocLD)
Test set for the FUNSD benchmark used with DocLD: 50 form images and 50 ground-truth annotation JSON files.
images/ — 50 PNG form images (FUNSD test set).
annotations/ — 50 JSON files (FUNSD format: entities with text, box, label).
Code and full benchmark: github.com/Doc-LD/funsd-bench.