Deep FCN for Arabic Scene Text Detection
Ines Beltaief, Mohamed Ben Halima · 2018
Visual text is considered as one of the major indispensable aspects of communication field used by individuals and broadly applied in our daily Transactions. Thus, detecting and exploiting this textual information is of a big prominence. State of the art methods for detecting text on printed documents has achieved impressing results on both accuracy and precision values thanks to the sophisticated deep earning approaches, while researchers on natural scenes images still on progress due to the various difficulties on distinguishing text candidates from the remaining shapes. wherefore, as a fast and efficient solution, we propose a deep incorporated multilingual scene text detector system to forthwith localize text using an end-to-end trainable single Network. For training and testing stages, we have used the ACTIV [24] dataset.