Datasets:
Tasks:
Image-Text-to-Text
Formats:
parquet
Languages:
Portuguese
Size:
< 1K
ArXiv:
Tags:
ocr
benchmark
document-understanding
brazilian-portuguese
text-recognition
handwriting-recognition
License:
Commit History
-
chore(deps): add ipykernel to requirements
a46b63c
-
fix(eval): replace unsafe eval() with json.loads in benchmark parser
The plain-text normaliser parsed the ground-truth JSON with eval() after a "null" -> "'null'" string replace. That executed arbitrary code from the dataset and silently mis-parsed native JSON null/true/false. Checked against the DharmaOCR-Benchmark test split: the assistant column is valid JSON, so json.loads parses it and drops null fields as intended, and no ast.literal_eval fallback is needed. The same diff also polishes the notebook (mechanical, no metric change): drop the dead run_inference wrapper and an unnecessary NLTK punkt download, align the max_tokens default with MAX_TOKENS (8192), and fix stale column/docstring text. benchmark_score and the Levenshtein ratio are unchanged.1819971
-
add evaluation notebook and frozen requirements
Verified signaturee8f4bb5 -
add evaluation notebook and frozen requirements
Verified signature43b3a45 -
fix(viewer): smaller row groups + page index (shard 11/11)
Verified signature851d544 -
fix(viewer): smaller row groups + page index (shard 10/11)
Verified signaturead2e17d -
fix(viewer): smaller row groups + page index (shard 9/11)
Verified signature618f7a0 -
fix(viewer): smaller row groups + page index (shard 8/11)
Verified signature4123c9e -
fix(viewer): smaller row groups + page index (shard 7/11)
Verified signatureb8a8f69 -
fix(viewer): smaller row groups + page index (shard 6/11)
Verified signature0693d55 -
fix(viewer): smaller row groups + page index (shard 5/11)
Verified signatureafb1dc9 -
fix(viewer): smaller row groups + page index (shard 4/11)
Verified signature1398527 -
fix(viewer): smaller row groups + page index (shard 3/11)
Verified signature051df8d -
fix(viewer): smaller row groups + page index (shard 2/11)
Verified signature94eb8cc -
fix(viewer): smaller row groups + page index (shard 1/11)
Verified signatured9cd510 -
fix(viewer): smaller row groups + page index (shard 0/11)
Verified signature01afec6 -
chore: remove oversized shards (will re-upload with smaller row groups)
Verified signature439a6cc -
chore: remove oversized shards (will re-upload with smaller row groups)
Verified signaturef0d4edb -
chore: remove oversized shards (will re-upload with smaller row groups)
Verified signature4627139 -
chore: remove oversized shards (will re-upload with smaller row groups)
Verified signature2cfc5f4 -
chore: remove oversized shards (will re-upload with smaller row groups)
Verified signaturef935935 -
chore: remove oversized shards (will re-upload with smaller row groups)
Verified signature02dc230 -
fix: correct assistant_without_json error rows (shard 5)
Verified signature418e9e3 -
fix: correct assistant_without_json error rows (shard 4)
Verified signaturebf55b07 -
fix: correct assistant_without_json error rows (shard 3)
Verified signaturea504695 -
fix: correct assistant_without_json error rows (shard 2)
Verified signature618ef91 -
fix: correct assistant_without_json error rows (shard 1)
Verified signature5f1a9ea -
fix: correct assistant_without_json error rows (shard 0)
Verified signature46ee142
-
Update README.md
Verified signature4446cdc
-
Restore README to commit dffb3c0554b2 (id column, test split)
Verified signatured66c66f -
Fix README YAML: single test split only, correct metadata
Verified signature6ed670b -
Fix README: revert to test split only
Verified signatureb3bf9b0 -
Remove benchmark shard: data/benchmark-00005-of-00006.parquet
Verified signature03d7a86 -
Remove benchmark shard: data/benchmark-00004-of-00006.parquet
Verified signaturea8c53e8 -
Remove benchmark shard: data/benchmark-00003-of-00006.parquet
Verified signature57ef47a -
Remove benchmark shard: data/benchmark-00002-of-00006.parquet
Verified signature08d2a28 -
Remove benchmark shard: data/benchmark-00001-of-00006.parquet
Verified signature0bc160e -
Remove benchmark shard: data/benchmark-00000-of-00006.parquet
Verified signature484db31 -
Revert split name back to test
Verified signature21362b3 -
Fix README: remove stale test split, keep only benchmark split
Verified signaturea7e8b82 -
Remove old test shard: data/test-00005-of-00006.parquet
Verified signature48e50c8 -
Remove old test shard: data/test-00004-of-00006.parquet
Verified signature566f9d8 -
Remove old test shard: data/test-00003-of-00006.parquet
Verified signaturef5259b0 -
Remove old test shard: data/test-00002-of-00006.parquet
Verified signature1005621 -
Remove old test shard: data/test-00001-of-00006.parquet
Verified signature7564eec -
Remove old test shard: data/test-00000-of-00006.parquet
Verified signaturebb57796 -
Rename split: test -> benchmark
Verified signature423de0c -
Add id column (incremental 0-495), Image() type fixed
Verified signaturedffb3c0 -
Remove old parquet file: data/dharmaocr_benchmark.parquet
Verified signature9f3d7c5 -
Remove old parquet file: data/dharmaocr-benchmark.parquet
Verified signature279bc85