ICDAR

ICDAR
Name	ICDAR
Status	Active
Genre	Academic conference
Frequency	Biennial (historically)
Country	International

ICDAR

ICDAR is an international conference series focused on document analysis, pattern recognition, optical character recognition, and related fields. It brings together researchers from institutions such as Massachusetts Institute of Technology, Stanford University, Carnegie Mellon University, University of Oxford, and Tsinghua University alongside industrial partners like Google, Microsoft, IBM, and Amazon. The event has strong connections with professional societies including IEEE, ACM, IAPR, and national laboratories such as Lawrence Berkeley National Laboratory and Los Alamos National Laboratory.

History

The origins trace to early workshops on optical character recognition and document understanding influenced by pioneers at Bell Labs, Hewlett-Packard, Xerox PARC, and research groups at University of Cambridge and École Polytechnique Fédérale de Lausanne. Over decades, the series evolved through collaborations with International Association for Pattern Recognition, regional venues like CVPR-adjacent meetings, and partnerships with conferences such as ECCV, ICML, NeurIPS, and ICLR. Influential figures from Yann LeCun, Geoffrey Hinton, Andrew Ng, Alex Graves, and Jürgen Schmidhuber shaped algorithmic trends that appeared at the conference. The conference timeline intersects with landmark events like the adoption of convolutional neural networks, the rise of sequence-to-sequence models popularized at ACL and EMNLP, and evaluation campaigns similar to those run by ImageNet and PASCAL VOC.

Scope and Topics

ICDAR covers technologies and applications linked to document imagery and text recognition used by organizations such as United Nations, World Bank, European Commission, and NASA. Core topics include optical character recognition techniques developed at labs like Google Research and Facebook AI Research, layout analysis connected to work from Adobe Systems and SAP, handwriting recognition with heritage from IBM Research and MIT CSAIL, and scene text detection related to studies at Alibaba Group and ByteDance. Cross-disciplinary links connect with datasets and challenges associated with MNIST, SVHN, COCO, and evaluation protocols influenced by NIST and ISO standards committees. Applications encompass document forensics used by Interpol and FBI, digital humanities projects at British Library and Library of Congress, and mobile OCR in products by Apple and Samsung.

Conference Structure and Activities

Typical programs mirror architectures seen in NeurIPS and ICML with keynote addresses by leading researchers from institutions such as ETH Zurich and University of California, Berkeley. Sessions include peer-reviewed papers, poster sessions resembling those at SIGGRAPH, tutorials comparable to KDD workshops, and special sessions akin to AAAI symposia. Industrial tracks feature demos from companies like Dropbox and Zebra Technologies, while workshops are organized in collaboration with groups at Peking University, National University of Singapore, and University of Tokyo. Evaluation workshops replicate benchmark practices from Kaggle competitions and cooperative initiatives with agencies such as DARPA and European Space Agency.

Notable Papers and Contributions

Papers presented have advanced techniques from hidden Markov models popularized in Speech Processing communities to deep learning architectures influenced by publications at CVPR and ICLR. Landmark contributions include advances in convolutional recurrent networks inspired by work from NYU and Google Brain, robust binarization methods reflecting research at Tsinghua University and University of Illinois Urbana-Champaign, and multilingual OCR systems with collaborations involving Microsoft Research Asia and Baidu Research. Influential evaluations have paralleled benchmarks like GLUE and SQuAD in driving progress, and methods have been adopted in products by Adobe and standards bodies like IEEE Standards Association.

Awards and Competitions

The conference hosts competitions modeled after challenges in ImageNet and Pascal VOC with tasks for text localization, recognition, and layout analysis. Award categories include best paper recognitions similar to honors at ICLR and career awards reflecting precedents from AAAI and IEEE Computer Society. Past winners have come from teams at Microsoft Research, Alibaba DAMO Academy, Google DeepMind, and universities including University of Michigan, Columbia University, and University of Sydney.

Organizing Bodies and Sponsors

Organization and sponsorship often involve collaborations among IAPR, IEEE Computer Society, national research councils such as National Science Foundation and Engineering and Physical Sciences Research Council, and corporate sponsors including Intel, NVIDIA, Oracle, and Siemens. Local organizing committees have been hosted by universities like University of Florence, Seoul National University, Universidad de Chile, and Indian Institute of Science.

Category:Computer vision conferences Category:Pattern recognition