Pioneer dataset and automatic recognition of Urdu handwritten characters using a deep autoencoder and convolutional neural network

摘要: Automatic recognition of Urdu handwritten digits and characters, is a challenging task. It has applications in postal address reading, bank’s cheque processing, digitization preservation manuscripts from old ages. While there exists significant work for automatic English characters other major languages the world, done language extremely insufficient. This paper two goals. Firstly, we introduce pioneer dataset Urdu, containing samples more than 900 individuals. Secondly, report results as achieved by using deep auto-encoder network convolutional neural network. More specifically, use two-layer three-layer autoencoder evaluate frameworks terms accuracy. The proposed framework can successfully recognize with an accuracy 97% only, 81% only 82% both simultaneously. In comparison, 96.7% 86.5% 82.7% These serve baselines future research on text.