{"success":true,"data":{"author":"IBM Research, Deep Search team (Pfitzmann, Auer, Dolfi, Nassar, Staar)","bucketAccessType":"public","bucketName":"docling-project/DocLayNet-v1.2","bucketPrefix":"","category":"Computer Vision","cloudProvider":"huggingface","createdAt":"2026-11-14T14:15:13.099Z","description":"DocLayNet is IBM Research's human-annotated benchmark for document layout analysis, published at KDD 2022. It gives page-by-page layout segmentation ground truth as bounding boxes and polygons for 11 classes (Caption, Footnote, Formula, List-item, Page-footer, Page-header, Picture, Section-header, Table, Text, Title) on 80,863 unique pages drawn from six document categories: financial reports, scientific articles, laws and regulations, government tenders, manuals and patents. Pages were labelled by trained annotators and a fraction are double- or triple-annotated, which lets users estimate annotation agreement as an upper bound on model accuracy. Version 1.2 extends the original release by embedding the source PDF of each page in a binary column alongside the page image, the bounding boxes, category ids, segmentation polygons, the PDF text cells that fall inside each box, and page and document metadata. Fixed train, validation and test splits avoid leakage of layout styles. Measured 2026-11-14: 85 Parquet shards (79 train-and-val, 6 test) in data/, about 39.8 GB in total. Open with the Hugging Face datasets library, pandas or pyarrow; it is the standard training set for layout detectors such as those used in Docling and for object detection models applied to PDFs. Licensed by IBM under the Community Data License Agreement Permissive 1.0; cite the KDD 2022 paper.","displayTitle":"DocLayNet v1.2: human-annotated document layout segmentation","doi":"10.1145/3534678.3539043","id":"bf0cacc0-5473-5c5d-b563-604275206c8d","license":"CDLA-Permissive-1.0","licenseEvidenceUrl":"https://github.com/DS4SD/DocLayNet/blob/main/LICENSE","objectCountEstimate":87,"probeStatus":"ok","referenceUrl":"https://doi.org/10.1145/3534678.3539043","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"docling-project/DocLayNet-v1.2\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"docling-project/DocLayNet-v1.2\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=docling-project, size_categories:10K<n<100K, format:parquet, downloads=5824"],"format":"mixed","generatedAt":"2026-11-14T14:15:17.413Z","layout":{"example":"data/test-00000-of-00006.parquet","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"data/test-00000-of-00006.parquet","truncatedAt":512}],"sampleObjects":["data/test-00000-of-00006.parquet","data/test-00001-of-00006.parquet","data/test-00002-of-00006.parquet","data/test-00003-of-00006.parquet","data/test-00004-of-00006.parquet"],"sidecars":[{"bytes":"---\ndataset_info:\n  features:\n  - name: image\n    dtype: image\n  - name: bboxes\n    sequence:\n      sequence: float64\n  - name: category_id\n    sequence: int64\n  - name: segmentation\n    sequence:\n      sequence:\n        sequence: float64\n  - name: area\n    sequence: float64\n  - name: pdf_cells\n    list:\n      list:\n      - name: bbox\n        sequence: float64\n      - name: font\n        struct:\n        - name: color\n          sequence: int64\n        - name: name\n          dtype: string\n        - name: size\n          dtype: float64\n      - name: text\n        dtype: string\n  - name: metadata\n    struct:\n    - name: coco_height\n      dtype: int64\n    - name: coco_width\n      dtype: int64\n    - name: collection\n      dtype: string\n    - name: doc_category\n      dtype: string\n    - name: image_id\n      dtype: int64\n    - name: num_pages\n      dtype: int64\n    - name: original_filename\n      dtype: string\n    - name: original_height\n      dtype: float64\n    - name: original_width\n      dtype: float64\n    - name: page_hash\n      dtype: string\n    - name: page_no\n      dtype: int64\n  - name: pdf\n    dtype: binary\n  - name: modalities\n    sequence: string\n  splits:\n  - name: train\n    num_bytes: 35626146180.25\n    num_examples: 69375\n  - name: validation\n    num_bytes: 3090589267.941\n    num_examples: 6489\n  - name: test\n    num_bytes: 2529339432.131\n    num_examples: 4999\n  download_size: 39770621829\n  dataset_size: 41246074880.322\nconfigs:\n- config_name: default\n  data_files:\n  - split: train\n    path: data/train-*\n  - split: validation\n    path: data/validation-*\n  - split: test\n    path: data/test-*\n---\n# Dataset Card for DocLayNet v1.2\n\n## Dataset Description\n\n- **Homepage:** https://developer.ibm.com/exchanges/data/all/doclaynet/\n- **Repository:** https://github.com/DS4SD/DocLayNet\n- **Paper:** https://doi.org/10.1145/3534678.3539043\n\n\n### Dataset Summary\n\nThis dataset is an extention of the [original DocLayNet dataset](https://github.com/DS4SD/DocLayNet) which embeds the PDF files of the document images inside a binary column.\n\nDocLayNet provides page-by-page layout segmentation ground-truth using bounding-boxes for 11 distinct class labels on 80863 unique pages from 6 document categories. It provides several unique features compared to related work such as PubLayNet or DocBank:\n\n1. *Human Annotation*: DocLayNet is hand-annotated by well-trained experts, providing a gold-standard in layout segmentation through human recognition and interpretation of each page layout\n2. *Large layout variability*: DocLayNet includes diverse and complex layouts from a large variety of public sources in Finance, Science, Patents, Tenders, Law texts and Manuals\n3. *Detailed label set*: DocLayNet defines 11 class labels to distinguish layout features in high detail.\n4. *Redundant annotations*: A fraction of the pages in DocLayNet are double- or triple-annotated, allowing to estimate annotation uncertainty and an upper-bound of achievable prediction accuracy with ML models\n5. *Pre-defined train- test- and validation-sets*: DocLayNet provides fixed sets for each to ensure proportional representation of the class-labels and avoid leakage of unique layout styles across the sets.\n\n\n## Dataset Structure\n\nThis dataset is structured differently from the other repository [ds4sd/DocLayNet](https://huggingface.co/datasets/ds4sd/DocLayNet), as this one includes the content (PDF cells) of the detections, and abandons the COCO format.\n\n* `image`: page PIL image.\n* `bboxes`: a list of layout bounding boxes.\n* `category_id`: a list of class ids corresponding to the bounding boxes.\n* `segmentation`: a list of layout segmentation polygons.\n* `area`: Area of the bboxes.\n* `pdf_cells`: a list of lists corresponding to `bbox`. Each list contains the PDF cells (content) inside the bbox.\n* `metadata`: page and document metadetails.\n* `pdf`: Binary blob with the original PDF image.\n\nBounding boxes classes / categories:\n\n```\n1: Caption\n2: Footnote\n3: Formula\n4: List-item\n5: Page-footer\n6: Page-header\n7: Picture\n8: Section-header\n9: Table\n10: Text\n11: Title\n```\n\n\nThe `[","encoding":"utf8","key":"README.md","name":"README.md","size":5995,"truncated":true}],"sourceObjectKey":"data/test-00000-of-00006.parquet","starterCell":"from datasets import load_dataset\nds = load_dataset(\"docling-project/DocLayNet-v1.2\")\nprint(ds)","starterSamplePath":"data/test-00000-of-00006.parquet","totalObjects":87},"summary":"DocLayNet is IBM Research's human-annotated benchmark for document layout analysis, published at KDD 2022.","title":"DocLayNet v1.2: human-annotated document layout segmentation","updatedAt":"2026-11-14T17:30:23.957Z","visibility":"public","websiteUrl":"https://github.com/DS4SD/DocLayNet"}}