CITlab ARGUS for Arabic Handwriting

In recent years, it has been shown that multidimensional recurrent neural networks (MDRNN) perform very well in offline handwriting recognition problems like the OpenHaRT 2013 Document Image Recognition (DIR) task. With suitable writing preprocessing and dictionary lookup, our ARGUS software completed this task with an error rate of 26.27% in its primary setup. Keywords—handwriting recognition, neural network, LSTM