Arabic Handwriting Recognition System Based on Genetic Algorithm and Deep CNN Architectures
Kamline Miloud, Abdelmounaïm Moulay Lakhdar, Bendjillali Ridha Ilyas
Abstract
It is vital to understand Arabic handwritten characters because of their numerous advantages and applications. This research paper proposes a non-segmented Arabic Handwritten Recognition (AHR) system. For the best results, the genetic algorithm was used to optimize the VGG-16 and ResNet-50 Deep Convolutional Neural Networks (CNN) architectures to extract features followed by the classification. Our investigation was based on the HACDB database. Finally., a comparison of the proposed method to other approaches demonstrates its efficacy and robustness. The optimized ResNet-50 architecture was able to achieve a 100% success rate.
Topics & Concepts
Computer scienceConvolutional neural networkRobustness (evolution)ArabicArtificial intelligencePattern recognition (psychology)HandwritingHandwriting recognitionResidual neural networkArchitectureGenetic algorithmSpeech recognitionFeature extractionMachine learningChemistryVisual artsPhilosophyBiochemistryLinguisticsArtGeneHandwritten Text Recognition TechniquesVehicle License Plate RecognitionAdvanced Neural Network Applications