International Conference on Document Analysis and Recognition

International Conference on Document Analysis and Recognition
Name	International Conference on Document Analysis and Recognition
Acronym	ICDAR
Established	1991
Discipline	Optical character recognition; pattern recognition; document analysis
Frequency	Biennial (initially triennial)

International Conference on Document Analysis and Recognition is an international scientific conference focused on optical character recognition, pattern recognition, and document image analysis with broad participation from researchers linked to IEEE, International Association for Pattern Recognition, University of Cambridge, Massachusetts Institute of Technology, Stanford University, and Carnegie Mellon University. The conference serves as a forum connecting developers from Google, Microsoft Research, IBM Research, and Adobe Systems with academics from institutions such as Tsinghua University, University of Oxford, ETH Zurich, National University of Singapore, and Peking University.

History and Background

ICDAR traces origins to early efforts in optical character recognition and handwriting recognition research driven by projects at Bell Labs, MIT Media Lab, IBM Watson Research Center, Los Alamos National Laboratory, and NASA Ames Research Center. Early meetings drew contributors influenced by results from the MNIST database, the NIST Special Database 19, and algorithmic advances like hidden Markov models, neural networks, support vector machines, and later convolutional neural networks. The conference timeline intersects with milestones exemplified by events such as the ImageNet challenge, the ICCV and CVPR conferences, and standards work at ISO committees, reflecting cross-pollination among groups at Academic Press, Springer, IEEE Computer Society, and major research laboratories.

Scope and Topics

The conference scope encompasses areas including handwriting recognition, document layout analysis, table recognition, form understanding, historical document processing, multilingual OCR, scene text recognition, document forensics, and information retrieval. Typical technical themes reference methods from deep learning, transformer (machine learning), graph neural networks, support vector machine, conditional random field, and hidden Markov model literature, with applications in projects from European Space Agency, World Bank, UNESCO, and cultural heritage institutions like the British Library and the Library of Congress.

Conference Organization and Governance

ICDAR is organized under sponsorship from entities such as the International Association for Pattern Recognition, the IEEE Signal Processing Society, and national committees including Chinese Academy of Sciences and French National Centre for Scientific Research. Governance typically involves an elected program committee chaired by researchers affiliated to University of Tokyo, University of California, Berkeley, Johns Hopkins University, Imperial College London, and University of Toronto. Program committees coordinate with local organizing committees comprising staff from host universities like University of Melbourne, Seoul National University, University of Sao Paulo, and EPFL.

Proceedings and Publications

Accepted papers are published in proceedings often distributed by IEEE Xplore and indexed in databases such as Scopus, Web of Science, DBLP, Google Scholar, and arXiv. Special issues drawing from ICDAR submissions appear in journals like IEEE Transactions on Pattern Analysis and Machine Intelligence, Pattern Recognition, International Journal on Document Analysis and Recognition, and Computer Vision and Image Understanding. Benchmark datasets and challenges introduced at ICDAR have been incorporated into repositories curated by groups at UC Irvine Machine Learning Repository, Kaggle, and GitHub organizations maintained by researchers from Facebook AI Research and OpenAI.

Notable Contributions and Awards

ICDAR has been the venue for seminal contributions such as datasets that advanced scene-text recognition paralleling breakthroughs reported at NeurIPS, algorithmic innovations influenced by LeNet and AlexNet, and evaluations that shaped standards similar to those from ISO/IEC. Award programs recognize papers and authors with accolades like the ICDAR Best Paper Award, Early Career Awards, and Best Dataset/Software awards, frequently honoring researchers associated with Yann LeCun, Geoffrey Hinton, Andrew Ng, Fei-Fei Li, and laboratories such as Microsoft Research Redmond and Google Research. Highly cited works from ICDAR have influenced deployments in products by ABBYY, Nuance Communications, and governmental digitization initiatives at National Archives and Records Administration and Bibliothèque nationale de France.

Hosting Venues and Frequency

Historically hosted at cities including Stuttgart, Utrecht, Seoul, Beijing, Montreal, Pisa, Beijing, Bangalore, Nara, Santiago, Beirut, Doha, and Sydney, ICDAR follows a cycle coordinated with regional committees from Asia-Pacific Artificial Intelligence Association, European Conference on Computer Vision affiliates, and North American organizers. The conference initially adopted a triennial rhythm before moving to a biennial schedule, and special annual editions have been convened to coincide with anniversaries and thematic collaborations with conferences like ICCV, ECCV, and CVPR.

Related Workshops and Special Sessions

ICDAR hosts numerous workshops and special sessions including focused events on handwritten text recognition competitions, historical document processing workshops, table recognition challenges, scene text understanding tutorials, and collaborative tracks with ICPR, ACL, SIGIR, and JCDL. Workshops often partner with initiatives like the READ (Recognition and Enrichment of Archival Documents) project, the PRImA Research Lab benchmarking activities, and challenge series similar to ROBUST and TextLoc. These sessions bring together stakeholders from European Union Horizon 2020 projects, cultural institutions such as the Vatican Library, and industry partners like Amazon Web Services and Intel Labs.

Category:Computer vision conferences Category:Pattern recognition conferences