Competitions | ICDAR 2015

Document Analysis Systems

The competition presents challenges for page segmentation, region classification, and text recognition in an end-to-end scenario. The dataset contains scanned pages from contemporary magazines and technical articles. Participants will be provided with know-how and tools that aid the development or extension of their page analysis systems.

“Robust Reading” refers to the interpretation of written communication in unconstrained settings. The ICDAR 2015 Robust Reading competition will build upon the success of the previous editions and will introduce an “end-to-end” task aiming at simultaneous word localisation and recognition in scene images, born-digital images and scene videos as well as a new large dataset (in the thousands of images) on incidental scene text.

Smartphones are replacing personal scanners. With more than 1.2 billion units sold in 2014, what was a trend is now an established use, and we all need reliable solutions for digitizing document images in a seamless way, to later search them, reuse their content, edit them, share them, and perform various other actions which we normally require on daily basis. This competition proposes 2 independant challenges in this topic: 1/ detect and segment page object in preview frames (to assist the user, and enable automated image enhancement); and 2/ extract and recognize text contained in mobile captured images (for indexation or editing purposes). Two new datasets will be released at this occasion.

The goal of this competition is to promote the Handwriting Text Recognition in historical handwritten documents. A subset of the Bentham manuscripts researched in the tranScriptorium project will be used. It has more than 80,000 documents, most of them digitised, and more than 6,000 have been transcribed with in a crowd-sourcing initiative. In this edition of the contest, 796 images will be used.

The aim of this competition is to evaluate the performance of algorithms for detecting lines of handwritten text in paragraph form drawn from historical documents. In particular, we wish to investigate and compare general methods that can reliably and robustly identify the origin point for text lines in the presence of various noise conditions, interfering annotations, and the artifacts common to historical documents. Specifically, we wish to only consider the task of finding the baseline of the first character of the left-most word of each line of text. The goal with this competition is to understand how this generality is achieved in hopes that it might open new avenues for consideration in other (more challenging) areas of layout analysis.

This competition aims to objectively compare different Keyword Spotting (KWS) approaches. In order to make the competition interesting for researchers from all backgrounds we consider: segmentation-free vs.segmentation-based, query-by-example vs. query-by-string and training-free vs. training-based systems, divided in two main tracks.All scenarios will be evaluated with the same data and metrics.

The MS-TEx contest aims to promote recent development in state-of-the-art methodologies directed to segmentation and binarization of the original text on multispectral document images. The advantage of multispectral images is that they offer an opportunity to achieve better differentiation between the different document image patterns.

Researchers and developers in the fields of signature verification and handwriting analysis are invited to participate in the SigWIComp2015. This will actually be a set of competitions with several tasks, covering different modalities (On- and Off-line), different scripts (Western and Indic), and different tasks (Signature Verification and Writer Identification). Participants are welcome to register for all, or any of the tasks they prefer.

This competition is aimed at writer identification and gender classification from offline handwritten documents using the QUWI database. The most interesting aspect is the dataset with writing samples of the same individual in Arabic as well as English allowing not only to objectively compare different systems but also to investigate the performance of traditional script-dependent systems in a multi-script experimental setup.

In multi-lingual and multi-script countries the use of two or more scripts is quite common for information communication through news and advertisement videos. The text present in videos plays an important role in automatic video indexing and retrieval, hence, OCR of multi-lingual video-text is crucial. The objective of the competition is to identify different scripts from the extracted video words.

This competition aims to motivate research around Text Images Super-Resolution (SR). Simple interpolation techniques are very limited at improving OCR performance. SR approaches aim to enhance the reconstruction process by generating missing details, yielding better OCR performances. Evaluation is based on both OCR accuracy and PSNR improvement compared with a simple bicubic interpolation of the LR images.

========================================

Call for Competitions <link>

April 30, 2015: Deadline for submission (for review) of full papers describing the competitions. Papers must be sent directly to the Competition Chairs. (extended to 5th May)
May 31, 2015: Review results on the papers.
June 14, 2015: Accepted camera-ready papers reporting on the competitions for inclusion in the proceedings.

========================================

Partners:	Sponsors: