ICDAR 2003

Competitions

Introduction
Datasets
Results
Cursive Script Recognition
Page Segmentation
Table Segmentation
Robust Reading
Robust Word Recognition
Robust Character Recognition
Text Locating
Sponsors
 
ICDAR2003 Logo

Introduction

New: ICDAR 2005 Competitions


 

The ICDAR 2003 competitions are now closed. 

The only competitions to receive any entries were Text Locating and Page Segmentation.  Papers describing the results of these competitions were published in the proceedings of ICDAR 2003, and are available on-line:

A more thorough write-up of the Text Locating results (including also some new results) has been submitted for publication in the IJDAR Special Issue on Reading in Camera Captured Scenes.

The rest of these pages have been left in tact for reference.


 

Welcome to the main competitions page for ICDAR 2003.

Phase 1

The first phase of the competitions is now closed.? During this first phase we received entries for the Page Segmentation and the Text Locating contests. Results will be presented in the Competitions Session of ICDAR 2003, and also published in the conference proceedings, and on this website after the conference.

Phase 2

Where there is sufficient interest, contests may be run in a second phase, with results published at or after the ICDAR conference.

The plan for Robust Reading contests (Robust OCR, Robust Word Recognition, Text Locating and Robust Reading) is to publish the results of these in a paper to appear in an IJDAR Special Issue on Reading in Camera Captured Scenes .? Winners of each contest will be invited to co-author a paper with details of their method.? (Note: these papers will be subject to the normal peer review process, and are not guaranteed to be published).

Entry: please email the competition organizer (cc to Simon Lucas ) if you are interested in entering any of these competitions.

Open Mode Competitions

The Page Segmentation and Table Segmentation competitions will be run in open mode . Contestants are responsible for running their algorithms on the datasets and generating their own outputs in a standard format. These raw results are then analyzed using some standard performance measure. The open mode allows researchers to experiment with and tune various algorithms, and supply the results that they choose. Please refer to the individual competition pages for more details.

Closed Mode Competitions

The other competitions will be run in closed mode . Contestants must supply executable code to process the competition input data and produce raw results in a standard format. These raw results are then analysed using some standard performance measure. The executables should run on a windows platform, unless otherwise stated or arranged. The closed mode measures the "out of the box" performance of a system i.e. how well it performs on unseen data when there is no researcher present to tune it.

Community: ? visit the discussion forum to share your views with the document analysis and recognition community.? Have your say on dataset formats, test image characteristics, software interfaces and evaluation procedures.





Hosted with kind thanks to the University of Essex , ? 2002.

University of Essex