ICDAR 2011 - French Handwriting Recognition Competition

Abstract

This paper describes the French handwriting recognition competition held at ICDAR 2011. This competition is based on the RIMES-database composed of French written documents corresponding to letters sent by individuals to companies or administrations. Two tasks have been proposed this year : the first one consists in recognizing isolated snippets of words with the help of a given dictionary; the second one consists in recognizing blocks of words segmented into lines. This year 9 systems were submitted for the different competition subtasks. A comparison between different classification and recognition systems show interesting results.A short description of the participating groups, their systems, and the results achieved are presented.

Steel defect classification with Max-Pooling Convolutional Neural Networks

“…To avoid this problem predictions of various classifiers trained on differently normalized data can be combined [27]. Along similar lines the same classifier can be trained on random subsets of the training set (bootstrap aggregation technique [28]), or different types of classifiers can be trained on the same data [29]. Here we combine classifiers trained on different features, harnessing the complementary information content of the various feature descriptors.…”

Comparative Study of HMM and BLSTM Segmentation-Free Approaches for the Recognition of Handwritten Text-Lines

“…Cluster minimal occupancy and likelihood threshold have both been optimized on a distinct validation database. This approach led to improved recognition performance as highlighted in [12] and this system participated in Arabic and Latin handwritten word recognition competitions [7], [8].…”

“…The scope of recognition systems has been enlarged from word recognition to text-line recognition from several perspectives [5], [6]. Our main motivation is based on the recent emergence of combined or hybrid recognizers, especially in international handwriting recognition competitions [7], [8]. Such systems are top ranked according to recognition performance.…”

Using the Web to Create Dynamic Dictionaries in Handwritten Out-of-Vocabulary Word Recognition

“…To assess the effectiveness of dynamic dictionary creation presented in Section IV, we conduct experiments on the RIMES word database [18]. The metric used throughout all experiments is accuracy (i.e.…”

“…From these documents, blocks, text-lines and words were extracted and labeled. Word and text-line recognition competitions have been organized using this database in 2009 and 2011 [18]. RIMES 2011 is divided into training, validation and test sets composed of 51,739, 7,464 and 7,776 word images respectively.…”