HJDataset is a large dataset of Historical Japanese Documents with Complex Layouts. It contains over 250,000 layout element annotations of seven types. In addition to bounding boxes and masks of the content regions, it also includes the hierarchical structures and reading orders for layout elements. The dataset is constructed using a combination of human and machine efforts. Source: A Large Dataset of Historical Japanese Documents with Complex Layouts Image Source: https://dell-research-harvard.github.io/HJDataset/

The IAM database contains 13,353 images of handwritten lines of text created by 657 writers. The texts those writers transcribed are from the Lancaster-Oslo/Bergen Corpus of British English. It includes contributions from 657 writers making a total of 1,539 handwritten pages comprising of 115,320 words and is categorized as part of modern collection. The database is labeled at the sentence, line, and word levels. Source: Measuring Human Perception to Improve Handwritten Document Transcription Image Source: https://fki.tic.heia-fr.ch/databases/iam-handwriting-database

A dataset for evaluating text classification, domain adaptation, and active learning models. The dataset consists of 22,660 documents (tweets) collected in 2018 and 2019. It spans across four domains: Alzheimer's, Parkinson's, Cancer, and Diabetes.

We collect, organize and open-source the large-scale multimodal instruction dataset, Infinity-MM, consisting of tens of millions of samples. Through quality filtering and deduplication, the dataset has high quality and diversity. We propose a synthetic data generation method based on open-source models and labeling system, using detailed image annotations and diverse question generation.

This dataset arises from the READ project (Horizon 2020). The dataset consists of a subset of documents from the Ratsprotokolle collection composed of minutes of the council meetings held from 1470 to 1805 (about 30.000 pages), which will be used in the READ project. This dataset is written in Early Modern German. The number of writers is unknown. Handwriting in this collection is complex enough to challenge the HTR software. The training dataset is composed of 400 pages; most of the pages consist of a single block with many difficulties for line detection and extraction. The ground-truth in this set is in PAGE format and it is provided annotated at line level in the PAGE files. The previous dataset is the same that is located at https://zenodo.org/record/218236#.WnLhaCHhBGF The new file includes the test set corresponding to the HTR competition held at ICFHR 2016 Toselli, A.H., Romero, V., Villegas, M., Vidal, E., & Sánchez, J.A. (2018). HTR Dataset ICFHR 2016 (Version 1.2.0) [Data set]. Zenodo. http://doi.org/10.5281/zenodo.1297399

Specially designed to evaluate active learning for video object detection in road scenes. Source: Temporal Coherence for Active Learning in Videos

Twitter100k is a large-scale dataset for weakly supervised cross-media retrieval. Source: Twitter100k: A Real-world Dataset for Weakly Supervised Cross-Media Retrieval Image Source: https://arxiv.org/pdf/1703.06618v1.pdf

Bentham manuscripts refers to a large set of documents that were written by the renowned English philosopher and reformer Jeremy Bentham (1748-1832). Volunteers of the Transcribe Bentham initiative transcribed this collection. Currently, >6 000 documents or > 25 000 pages have been transcribed using this public web platform. For our experiments, we used the BenthamR0 dataset a part of the Bentham manuscripts.

The ChineseLP dataset contains 411 vehicle images (mostly of passenger cars) with Chinese license plates (LPs). It consists of 252 images captured by the authors and 159 images downloaded from the internet. The images present great variations in resolution (from 143 × 107 to 2048 × 1536 pixels), illumination and background.

COMP6 is a benchmark for evaluating the extensibility of machine-learning based molecular potentials. It contains a diverse set of organic molecules. Source: https://github.com/isayev/COMP6

The DDI-100 dataset is a synthetic dataset for text detection and recognition based on 7000 real unique document pages and consists of more than 100000 augmented images. The ground truth comprises text and stamp masks, text and characters bounding boxes with relevant annotations. Source: DDI-100

The database is written in Cyrillic and shares the same 33 characters. Besides these characters, the Kazakh alphabet also contains 9 additional specific characters. This dataset is a collection of forms. The sources of all the forms in the datasets were generated by LATEX which subsequently was filled out by persons with their handwriting. The database consists of more than 1400 filled forms. There are approximately 63000 sentences, more than 715699 symbols produced by approximately 200 diferent writers. We utilized three different datasets described as following: Handwritten samples (Forms) of keywords in Kazakh and Russian (Areas, Cities , Village , etc.) Handwritten Kazakh and Russian alphabet in cyrillic Handwritten samples (Forms) of poems in Russian Image source: https://github.com/abdoelsayed2016/HKR_Dataset

Handwritten Text Recognition (HTR) is an open problem at the intersection of Computer Vision and Natural Language Processing. The main challenges, when dealing with historical manuscripts, are due to the preservation of the paper support, the variability of the handwriting – even of the same author over a wide time-span – and the scarcity of data from ancient, poorly represented languages. With the aim of fostering the research on this topic, in this paper we present the Ludovico Antonio Muratori (LAM) dataset, a large line-level HTR dataset of Italian ancient manuscripts edited by a single author over 60 years. The dataset comes in two configurations: a basic splitting and a date-based splitting which takes into account the age of the author. The first setting is intended to study HTR on ancient documents in Italian, while the second focuses on the ability of HTR systems to recognize text written by the same writer in time periods for which training data are not available. For both configurations, we analyze quantitative and qualitative characteristics, also with respect to other line-level HTR benchmarks, and present the recognition performance of state-of-the-art HTR architectures. The dataset is available for download at https://aimagelab.ing.unimore.it/go/lam.

Arxiv GR-QC (General Relativity and Quantum Cosmology) collaboration network is from the e-print arXiv and covers scientific collaborations between authors papers submitted to General Relativity and Quantum Cosmology category. If an author i co-authored a paper with author j, the graph contains a undirected edge from i to j. If the paper is co-authored by k authors this generates a completely connected (sub)graph on k nodes. Source: https://snap.stanford.edu/data/ca-GrQc.html

This dataset contains Bangla handwritten numerals, basic characters and compound characters. This dataset was collected from multiple geographical location within Bangladesh and includes sample collected from a variety of aged groups. This dataset can also be used for other classification problems i.e: gender, age, district. Source: BanglaLekha-Isolated: A Comprehensive Bangla Handwritten Character Dataset

The BRUSH dataset (BRown University Stylus Handwriting) contains 27,649 online handwriting samples from a total of 170 writers. Every sequence is labeled with intended characters such that dataset users can identify to which character a point in a sequence corresponds. The dataset was introduced in the paper "Generating Handwriting via Decoupled Style Descriptors" by Atsunobu Kotani, Stefanie Tellex, James Tompkin from Brown University, presented at European Conference on Computer Vision (ECCV) 2020. ##### Terms of Use The BRUSH dataset may only be used for non-commercial research purposes. Anyone wanting to use it for other purposes should contact Prof. James Tompkin. If you publish materials based on this database, we request that you please include a reference to our paper. @inproceedings{kotani2020generating, title={Generating Handwriting via Decoupled Style Descriptors}, author={Kotani, Atsunobu and Tellex, Stefanie and Tompkin, James}, booktitle={European Conference on Computer Vision}, pages={764--780}, year={2020}, organization={Springer} }

Digital Peter is a dataset of Peter the Great's manuscripts annotated for segmentation and text recognition. The dataset may be useful for researchers to train handwriting text recognition models as a benchmark for comparing different models. It consists of 9,694 images and text files corresponding to lines in historical documents. The dataset includes Peter’s handwritten materials covering the period from 1709 to 1713. The open machine learning competition Digital Peter was held based on the considered dataset.

Arabic handwriting dataset.

Goldfinch is a dataset for fine-grained recognition challenges. It contains a list of bird, butterfly, aircraft, and dog categories with relevant Google image search and Flickr search URLs. In addition, it also includes a set of active learning annotations on dog categories. Source: The Unreasonable Effectiveness of Noisy Data for Fine-Grained Recognition

dataset link : https://www.kaggle.com/datasets/osamahosamabdellatif/high-quality-invoice-images-for-ocr Overview High-Quality Invoice Images for OCR is a curated dataset containing professionally scanned and digitally captured invoice documents. It is designed for training, fine-tuning, and evaluating OCR models, machine learning pipelines, and data extraction systems. This dataset focuses on clean, structured invoices to simulate real-world scenarios in financial document automation. What's Inside 📄 Variety of invoice templates from multiple industries (e.g., retail, manufacturing, services) 🖋️ Different currencies, tax formats, and layouts 📸 High-resolution scanned and photographed invoices 🏷️ Optional field annotations (e.g., invoice number, date, total amount, vendor name) for supervised training Key Applications Training and fine-tuning OCR and Document AI models Machine learning for structured and semi-structured data extraction Intelligent Document Processing (IDP) and Robotic Process Automation (RPA) Benchmarking table detection, key-value extraction, and layout analysis models Why Use This Dataset? ✅ High-quality images optimized for OCR and data extraction tasks ✅ Real-world invoice variations to improve model robustness ✅ Ideal for machine learning workflows in finance, ERP, and accounting systems ✅ Supports rapid prototyping for invoice understanding models Ideal For Researchers working on OCR and document understanding Developers building invoice processing systems Machine learning engineers fine-tuning models for data extraction Startups and enterprises automating financial workflows

Imgur5k is a large-scale handwritten in-the-wild dataset, containing challenging real world handwritten samples from nearly 5K writers. It consists of ~135K handwritten English words from 5K different images. As opposed to existing dataests for OCR which have limited variability in their images, the images in Imgur5K contain a diverse set of styles.

Kazakh offline Handwritten Text dataset (KOHTD) has 3000 handwritten exam papers and more than 140335 segmented images and there are approximately 922010 symbols. It can serve researchers in the field of handwriting recognition tasks by using deep and machine learning. Image source: https://github.com/abdoelsayed2016/KOHTD

Konzil dataset was created by specialists of the University of Greifswald. It contains manuscripts written in modern German. Train sample consists of 353 lines, validation - 29 lines and test - 87 lines.

The largest dataset of extracted visual content from historic newspapers ever produced. The Newspaper Navigator dataset, finetuned visual content recognition model. Source: The Newspaper Navigator Dataset: Extracting And Analyzing Visual Content from 16 Million Historic Newspaper Pages in Chronicling America

Patzig contains handwritten texts written in modern German. Train sample consists of 485 lines, validation - 38 lines and test -118 lines.

A benchmark for molecular machine learning where improvements in model performance can be immediately observed in the throughput of promising molecules synthesized in the lab. Photoswitches are a versatile class of molecule for medical and renewable energy applications where a molecule's efficacy is governed by its electronic transition wavelengths. Source: The Photoswitch Dataset: A Molecular Machine Learning Benchmark for the Advancement of Synthetic Chemistry

Ricordi contains handwritten texts written in Italian. Train sample consists of 295 lines, validation - 19 lines and test - 69 lines.

Schiller contains handwritten texts written in modern German. Train sample consists of 244 lines, validation - 21 lines and test - 63 lines.

Schwerin contains handwritten texts written in medieval German. Train sample consists of 793 lines, validation - 68 lines and test - 196 lines.

We introduce a new Dataset (BN-HTRd) for offline Handwritten Text Recognition (HTR) from images of Bangla scripts comprising words, lines, and document-level annotations. The BN-HTRd dataset is based on the BBC Bangla News corpus - which acted as ground truth texts for the handwritings. Our dataset contains a total of 786 full-page images collected from 150 different writers. With a staggering 1,08,181 instances of handwritten words, distributed over 14,383 lines and 23,115 unique words, this is currently the 'largest and most comprehensive dataset' in this field. We also provided the bounding box annotations (YOLO format) for the segmentation of words/lines and the ground truth annotations for full-text, along with the segmented images and their positions. The contents of our dataset came from a diverse news category, and annotators of different ages, genders, and backgrounds, having variability in writing styles. The BN-HTRd dataset can be adopted as a basis for various handwriting classification tasks such as end-to-end document recognition, word-spotting, word/line segmentation, and so on. The statistics of the original dataset are given below: - Number of writers = 150 - Total number of images = 786 - Total number of lines = 14,383 - Total number of words = 1,08,181 - Total number of unique words = 23,115 - Total number of punctuation = 7,446 - Total number of characters = 5,74,203 From v3.0 onwards, we are also providing automatic bounding box annotations (YOLO format) of 805 document images containing words/lines. The statistics of the automatic annotations are given below: - Number of writers = 87 - Total number of images = 805 - Total number of lines = 14,836 - Total number of words = 1,06,135

The L-Bird (Large-Bird) dataset contains nearly 4.8 million images which are obtained by searching images of a total of 10,982 bird species from the Internet. Source: Fine-Grained Visual Categorization using Meta-Learning Optimization with Sample Selection of Auxiliary Data Image Source: https://arxiv.org/pdf/1511.06789.pdf

MCSCSet is a large-scale specialist-annotated dataset, designed for the task of Medical-domain Chinese Spelling Correction that contains about 200k samples. MCSCSet involves: i) extensive real-world medical queries collected from Tencent Yidian, ii) corresponding misspelled sentences manually annotated by medical specialists. Source: MCSCSet: A Specialist-annotated Dataset for Medical-domain Chinese Spelling Correction Image Source: https://arxiv.org/pdf/2210.11720v1.pdf

A large scale OCSR dataset, proposed in paper “MolParser: End-to-end Visual Recognition of Molecule Structures in the Wild“ MolParser-7M contains nearly 8 million paired image-SMILES data. It should be noted that the caption of image is extended-SMILES format proposed in paper.

• 5 domains: synthetic domain, document domain, street view domain, handwritten domain, and car license domain * over five million images

Saint Gall dataset contains handwritten historical manuscripts written in Latin that date back to the 9th century. It consists of 60 pages, 1 410 text lines and 11 597 words.

Description We propose a new database for information extraction from historical handwritten documents. The corpus includes 5,393 finding aids from six different series, dating from the 18th-20th centuries. Finding aids are handwritten documents that contain metadata describing older archives. They are stored in the National Archives of France and are used by archivists to identify and find archival documents. Each document is annotated at page-level, and contains seven fields to retrieve. The localization of each field is not available in such a way that this dataset encourages research on segmentation-free systems for information extraction. The dataset is available at https://zenodo.org/record/7868059 ## Details for each series and entity type | Series | Train | Validation | Test | Total (%) | | ------------------- | ----- | ---------- | ---- | --------: | | E series | 322 | 64 | 79 | 8.6 | | L series | 38 | 8 | 4 | 0.9 | | M series | 128 | 21 | 27 | 3.3 | | X1a series | 2209 | 491 | 469 | 58.8 | | Y series | 940 | 205 | 196 | 24.9 | | Douët s'Arcq series | 141 | 22 | 29 | 3.5 | | Total | 3778 | 811 | 804 | 100 | | Entities | Train | Validation | Test | Total (%) | | -------------- | ----- | ---------- | ----- | --------: | | date | 8406 | 1814 | 1799 | 10.4 | | title | 35531 | 7495 | 8173 | 44.5 | | serie | 3168 | 664 | 676 | 3.9 | | analysis | 25988 | 5130 | 5602 | 31.9 | | volume_number | 3913 | 808 | 813 | 4.8 | | article_number | 3181 | 665 | 678 | 3.9 | | arrangement | 644 | 122 | 153 | 0.8 | | Total | 80831 | 16698 | 17894 | 100 | ## Data encoding Transcriptions with entities are encoded in the labels.json JSON file. Special tokens are used to represent named entities. Please not that there are only opening NER tokens: each entity spans all words until the next entity starts. | Entities | Special token | Symbol unicode | | -------------- | ------------- | -------- | | date | ⓓ | \u24d3 | | title | ⓘ | \u24d8 | | serie | ⓢ | \u24e2 | | analysis | ⓒ | \u24d2 | | volume_number | ⓟ | \u24df | | article_number | ⓐ | \u24d0 | | arrangement | ⓥ | \u24e5 | # Cite us! The dataset is presented in details in the following article: bib @article{simara2023, author = {Solène Tarride and Mélodie Boillet and Jean-François Moufflet and Christopher Kermorvant}, title = {SIMARA: a database for key-value information extraction from full-page handwritten documents}, year = {2023}, journal={Proceedings of the 17th International Conference on Document Analysis and Recognition}, }