Datasets:

bsmock
/

pubtables-1m

License:

Dataset Viewer

The dataset viewer is not available for this subset.

Cannot get the split names for the config 'default' of the dataset.

Exception:    SplitsNotFoundError
Message:      The split names could not be parsed from the dataset config.
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
                  for split_generator in builder._split_generators(
                                         ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 81, in _split_generators
                  first_examples = list(islice(pipeline, self.NUM_EXAMPLES_FOR_FEATURES_INFERENCE))
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 32, in _get_pipeline_from_tar
                  fs: fsspec.AbstractFileSystem = fsspec.filesystem("memory")
                                                  ~~~~~~~~~~~~~~~~~^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/fsspec/registry.py", line 302, in filesystem
                  cls = get_filesystem_class(protocol)
                File "/usr/local/lib/python3.14/site-packages/fsspec/registry.py", line 239, in get_filesystem_class
                  raise ValueError(f"Protocol not known: {protocol}")
              ValueError: Protocol not known: memory
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 66, in compute_split_names_from_streaming_response
                  for split in get_dataset_split_names(
                               ~~~~~~~~~~~~~~~~~~~~~~~^
                      path=dataset,
                      ^^^^^^^^^^^^^
                      config_name=config,
                      ^^^^^^^^^^^^^^^^^^^
                      token=hf_token,
                      ^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
                  info = get_dataset_config_info(
                      path,
                  ...<6 lines>...
                      **config_kwargs,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
                  raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
              datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

PubTables-1M

GitHub: https://github.com/microsoft/table-transformer
Paper: "PubTables-1M: Towards comprehensive table extraction from unstructured documents"
Hugging Face:
- Detection model
- Structure recognition model

Currently we only support downloading the dataset as tar.gz files. Integrating with HuggingFace Datasets is something we hope to support in the future!

Please switch to the "Files and versions" tab to download all of the files or use a command such as wget to download from the command line.

Once downloaded, use the included script "extract_structure_dataset.sh" to extract and organize all of the data.

Files

It comes in 18 tar.gz files:

Training and evaluation data for the structure recognition model (947,642 total cropped table instances):

PubTables-1M-Structure_Filelists.tar.gz
PubTables-1M-Structure_Annotations_Test.tar.gz: 93,834 XML files containing bounding boxes in PASCAL VOC format
PubTables-1M-Structure_Annotations_Train.tar.gz: 758,849 XML files containing bounding boxes in PASCAL VOC format
PubTables-1M-Structure_Annotations_Val.tar.gz: 94,959 XML files containing bounding boxes in PASCAL VOC format
PubTables-1M-Structure_Images_Test.tar.gz
PubTables-1M-Structure_Images_Train.tar.gz
PubTables-1M-Structure_Images_Val.tar.gz
PubTables-1M-Structure_Table_Words.tar.gz: Bounding boxes and text content for all of the words in each cropped table image

Training and evaluation data for the detection model (575,305 total document page instances):

PubTables-1M-Detection_Filelists.tar.gz
PubTables-1M-Detection_Annotations_Test.tar.gz: 57,125 XML files containing bounding boxes in PASCAL VOC format
PubTables-1M-Detection_Annotations_Train.tar.gz: 460,589 XML files containing bounding boxes in PASCAL VOC format
PubTables-1M-Detection_Annotations_Val.tar.gz: 57,591 XML files containing bounding boxes in PASCAL VOC format
PubTables-1M-Detection_Images_Test.tar.gz
PubTables-1M-Detection_Images_Train_Part1.tar.gz
PubTables-1M-Detection_Images_Train_Part2.tar.gz
PubTables-1M-Detection_Images_Val.tar.gz
PubTables-1M-Detection_Page_Words.tar.gz: Bounding boxes and text content for all of the words in each page image (plus some unused files)

Full table annotations for the source PDF files:

PubTables-1M-PDF_Annotations.tar.gz: Detailed annotations for all of the tables appearing in the source PubMed PDFs. All annotations are in PDF coordinates.
- 401,733 JSON files, one per source PDF document

Downloads last month: 1,799

Total file size:

117 GB

Models trained or fine-tuned on bsmock/pubtables-1m