Deep Learning Based Pashto Characters Recognition

In artificial intelligence, text identification and analysis that are based on images play a vital role in the retrieving process. Automatic recognition system development is difficult task machine learning, but case of cursive languages, it poses big challenge to research community due slight changes character’s shapes unavailability standard dataset. While this becomes more challenging Pashto language large number characters its dataset than other similar languages (Persian, Urdu, Arabic) change shape. This paper aims address accept these challenges by developing an optimal optical character (OCR) recognise isolated handwritten characters. The proposed OCR developed using multiple long short-term memory (LSTM) deep learning model. applicability model validated decision trees (DT) classification tool zoning feature extraction technique invariant moment approaches. An overall accuracy rate 89.03% calculated for LSTM-based while DT-based 72.9% achieved vector 74.56% moments-based map. Applicability evaluated different performance metrics accuracy, f-score, specificity, varying training test sets.