Exploitation de l’échelle d’écriture pour améliorer la reconnaissance automatique des textes manuscrits arabes

Chammas, Edgard; Mokbel, Chafic; Likforman-Sulem, Laurence

doi:10.3166/dn.19.2-3.95-115

Cited by 4 publications

(1 citation statement)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To further enhance the performance of the system, we exploited the variability in the writing scale to augment the training set with text-line images at multiple scales. Based on a vertical scale score [19], the training lines were first classified into 3 classes (Large, Medium and Small) via Jenks natural breaks optimization algorithm [20]. By dividing the training set over the three classes, the data volume per class become smaller.…”

Section: Integration Of Multi-scale Training Datamentioning

confidence: 99%

Handwriting Recognition of Historical Documents with Few Labeled Data

Chammas

Mokbel

Likforman-Sulem

2018

2018 13th IAPR International Workshop on Document Analysis Systems (DAS)

Self Cite

View full text Add to dashboard Cite

Historical documents present many challenges for offline handwriting recognition systems, among them, the segmentation and labeling steps. Carefully annotated textlines are needed to train an HTR system. In some scenarios, transcripts are only available at the paragraph level with no text-line information. In this work, we demonstrate how to train an HTR system with few labeled data. Specifically, we train a deep convolutional recurrent neural network (CRNN) system on only 10% of manually labeled text-line data from a dataset and propose an incremental training procedure that covers the rest of the data. Performance is further increased by augmenting the training set with specially crafted multiscale data. We also propose a model-based normalization scheme which considers the variability in the writing scale at the recognition phase. We apply this approach to the publicly available READ dataset 1 . Our system achieved the second best result during the ICDAR2017 competition [1].

show abstract

Section: Integration Of Multi-scale Training Datamentioning

confidence: 99%